Current food systems are challenged by an increasing demand to deliver sufficient and nutritious foods while simultaneously decreasing environmental impact (1, 2). Although it is possible to produce enough food globally to provide every person approximately 2,700 kcal [average requirement energy for men is 2,900 and 2,200 kcal for women (3)], more than 800 million people still suffer from chronic undernutrition (4). In addition, hidden hunger such as iron, vitamin A and zinc deficiencies still exists when many diets with sufficient calories do not provide adequate amounts of micronutrients (5, 6). Aside from health, diet has an impact on the environment through production and transport of foods (7, 8). Food systems contribute up to third of total global anthropogenic greenhouse gas emissions (GHGE) (9) and agriculture is responsible for 80–86% of food systems emissions and 70% of all water use (10, 11). The livestock sector contributes to 5.6–7.5 Gt CO₂-eq. of human-induced GHGE from 1995 to 2005 (12). A requirement for a healthy and sustainable food system is to provide foods that are not only beneficial to a healthy diet but also limit harmful environmental impacts, i.e., to balance between animal-based sources and plant-based sources.

Although there is no universally agreed set of indicators to measure the quality of diets, it can be estimated by assessing two main dimensions: adequacy of energy and nutrients; and food diversity (13–15). Currently, the probability approach is the most accurate method to assess the adequacy of nutrient intakes (16) by comparing the nutrient requirement and nutrient intake distributions. The probability of inadequate intake would naturally be very low for those with higher nutrient intakes and would be higher for those with lower nutrient intakes. This method requires the mean and shape of the requirement distribution for a particular nutrient which is only obtained from the quantitative dietary intake survey data (16). But this approach needs accurate quantitative dietary intake data which are often not available due to high costs, time demand, and low capacity available for data collection and analysis in low and middle-income countries. Therefore, a more simple dietary diversity score (DDS), validated as a proxy for micronutrient adequacy (15) is often used, not only to assess food diversity but also diet quality.

Following the 1986 “Doi Moi” or economic reform in Vietnam, when the government promoted food production of especially rice and livestock, and moved to a more liberal market (5, 6), the economy developed exponentially (GDP and foreign investment increased). Accordingly, the diets of Vietnamese people began to gradually include a greater proportion of animal products (17–19), with changes happening faster in urban and high-income households than in rural and low-income households (19).

Several studies examined the greenhouse gas emission (GHGE) and blue water use (BWU) (the volume of surface and groundwater was withdrawn to produce the product) related to food production and consumption (20–23). In Vietnam, half of the total GHGE (from all sectors) comes from agriculture, which includes rice farming, raising livestock, emission from arable land, and burning of agricultural products (24). However, high-quality data on this is lacking. Although Heller et al. (25) found that the per capita diet-related GHGE increased in Vietnam over the period 1971–2011, mainly associated with an increase in meat consumption, this was derived using FAO food balance sheets and not data collected with methods of a higher accuracy (e.g., 24-h recall, food frequency questionnaire; FFQ) (25). One study did utilize a higher quality data collection method (namely, 24-h recall), however, they only did so in northern Vietnam, meaning these results are not nationally representative (26). In order to amend this knowledge gap, this paper aimed to use the nationally representative General Nutrition Survey 2009–2010 to identify common dietary patterns in Vietnam, quantify their GHGEs and BWU, and compare their diet quality and environmental impact to average diets to evaluate potential trade-offs between healthy diets and environmental sustainability objectives.

We identified three distinct dietary patterns: an Omnivorous pattern with a high consumption of red meat, a rice-based pattern with characteristics of the traditional Vietnamese diets (Traditional pattern), and a Pescatarian pattern in which fish was mainly consumed. The Omnivorous pattern was predominantly consumed in the plateaus and by the wealthiest strata of the population, while the Traditional pattern was more often found in the mountainous areas and in populations with the lowest wealth. The Pescatarian pattern was more prevalent in the coastal region, while its association with wealth was weaker. Our results show that diet quality in terms of nutrient adequacy and diet diversity of Vietnamese women was low and varied according to adherence to the dietary patterns. The environmental impact of diets was higher than the 2050 target according to the work of de Pee et al. (49) as adapted from Willett et al. (2). Across the total population, estimated GHGE per 2,000 kcal was 4.82 kg CO₂-eq. per 2,000 kcal and blue water use was 0.12 m³ per 2,000 kcal and was highest in the Omnivorous pattern.

While the richer populations adhered more often to the Omnivorous pattern, the poorer populations adhered most to the Traditional pattern. The transition from a rice-based diet to a diet higher in consumption of red meat and sweet foods in the Omnivorous pattern reflects Westernization of the dietary intake in the richer populations of the delta regions. A similar so-called nutritional transition pattern and a staple diet pattern (comparable to the Omnivorous and Traditional patterns, respectively, identified in this study) were found in studies in Central Java, Indonesia, and northeast Thailand (50, 51).

In the present study, the average diets of Vietnamese women showed a low diversity (mean DDS was 4.4) and a high risk of micronutrient inadequacy (mean MPA was 38%). The mean DDS was lower than observed in China (5.0) (52). As for the MPA in the total population, this was higher than results from studies conducted in Bangladesh (35%) and Philippines (34%) in 2011 (47) and in China (30%) using the same set of nutrients (52). The low MPAs in both the total population and across all the dietary patterns reflect the high prevalence of micronutrient deficiency in Vietnam. According to National Institute of Nutrition, in 2015, on average, the prevalence of anemia was 25.5%, iron deficiency was 37.7%, and zinc deficiency was 63.6% in Vietnamese women (53).

To limit the environmental impact of various diets, global recommendations for GHGE emissions exist, such as provided by the EAT-Lancet which suggests a target of 1.36 kg CO₂-eq. (2, 49). The findings of our study showed that the average diet of the total population (4.82 kg CO₂-eq.) exceeded this recommendation. The mean GHGE of the Vietnamese diet was also higher compared to other studies in Vietnam, such as those of Heller et al. (approximately 3.33 CO₂-eq. per person) and Trinh et al. (3.17 kg CO₂-eq. per person) (25, 26). These differences may be explained by the use of different methods used in calculations. While we applied Clune’s dataset (54), which considered the full value chain from farm (production) to fork (consumption), the dataset of Heller et al. (which was also used by Trinh et al.) calculated the GHGE of food to the farm gate level, leaving transportation and consumption unaccounted for. The GHGE in our study was also higher than those found in China (2.9 kg CO₂-eq. per person) using a dataset from the same period, but lower than those reported for certain other regions with comparable diets, such as Hong Kong (5.7 kg CO₂-eq. per person per day) (55). The GHGE from our Vietnamese study were also lower than those of high income countries, such as the Netherlands, Chile, and Japan (56, 57).

There were significant differences of environmental impacts between Q1 and Q5 across the three dietary patterns. The Omnivorous pattern was characterized by a high intake of red meat, explaining the high GHGE of this pattern compared to the other two patterns. This corroborates with the finding of Rosi et al., in that the Omnivorous pattern in their study had a daily carbon emission 1.5x higher (4.0 kg CO₂-eq. per day) than the ovo-lacto-vegetarian diet (2.5 kg CO₂-eq.) (58). Blue water use also significantly differed between the patterns and the quintiles. For both the Omnivorous and Pescatarian patterns significant differences existed between Q1 and Q5, which can be owed to differential intake of rice and fruits, two food groups that require larger quantities of blue water to be cultivated. These findings, too, align with existing results, such as those of Kim et al. studying the contribution of plant foods and aquatic animals in the Pescatarian diet in India and Indonesia (57).

At the level of specific foods or food groups differences existed, too. Rice also contributed the most to the GHGE and blue water use in two of three dietary patterns (Traditional and Pescatarian pattern). In our study, the starchy staples, especially rice, contributed the most to the blue water footprint. The findings on blue water footprint were similar to a study in Vietnam from Trinh et al. but lower than a study from China (1.8 m³) (26, 55).

Overall, neither of the recommendations for health or those for environmental impact set by the EAT-Lancet were met by any of the three derived dietary patterns at the highest level of adherence (Q5). Even the Omnivorous pattern, which performed best from the perspective of diet quality, was still far from the EAT-Lancet recommendations (2, 49), and had the highest GHGE of all the patterns. Although GHGE was lower in the Traditional pattern, diet quality was the lowest of all the patterns. The Pescatarian pattern showed a marginally lower DDS than the omnivorous and an equivalent MPA; but in terms of environmental impact, the Pescatarian pattern performed better than the Omnivorous pattern but not as well as the Traditional pattern.

The findings of dietary patterns and their distributions within the regions suggest a role for targeted resource allocation in policy and government in improving diet quality and reducing environmental impact of diet by using local food-based strategies to be suitable for food and social norms in each region. This could be manifested by the implementation strategies on communication such as nutritional counseling and healthy diet advertisement campaigns in the urban and suburban areas, where consumption of meat and sweet foods were higher. In contrast, in the mountainous and poor communities, policies to improve the micronutrient status and to increase dietary diversity should be prioritized by for example promoting or facilitating home gardening, small animal husbandry, poultry and social marketing (59). In addition, the GHGEs we found were higher than recommended (EAT-Lancet commission) but were still lower than other high-income countries. National policies, including the National Nutrition Strategy 2010–2020, have identified necessary measures to improve the health status of the Vietnamese, however failed to consider the environmental impact of the diet and the importance of regional differences in dietary patterns. Results of the present study may help to guide a targeted approach to improve toward healthy and environmentally sustainable diets for everyone, everywhere in Vietnam. Thus, major changes in food systems in Vietnam will be required in order to allow for sustainable diets that are both healthier and reduce environmental impact.

Several limitations of this paper are acknowledged. Firstly, there were shortcomings in the dataset. The intake data used in the study was from 2010 and might not be representative of the current overall food intake. Nevertheless, the GNS was selected due to its availability, a large sample size and because of the nationally and regionally representative nature, meaning mean values are more likely to be accurate and outliers are better detected. Moreover, these advantages of the GNS provide a more complete picture of diet quality on the national and regional levels, although the aforementioned shortcomings should be considered. Secondly, at the time of data collection, the method of collection of food intake data was a single 24 h recall and at the household level, which is unable to estimate the usual intake of the individual. Despite the AFE method was validated and used to convert the data from household-level food consumption into individual-level intakes, the intake of protein and iron may be overestimated in adult women if the intra-household allocation of these nutrients was not equitable (31). The AFE method is based on the assumption that food is distributed in the household based on energy requirement. However, we do acknowledge this often is not the case, and distribution of food is often according to status in the household (31), meaning that the distribution for women is not equitable (or lower) than expected when based on their energy requirements. Furthermore, due to the limitation of this method, this study could not examine the individual characteristics (educational level, occupations) of these derived dietary patterns. Thirdly, at the time of this research, there were no Vietnam-specific GHGE or water footprint datasets available, therefore we selected the next most appropriate existing dataset to generate the environmental imprints. The environmental impact datasets were based on global estimations which will likely differ from the actual situation of Vietnam. By using external GHGE and BWU datasets, it may thus not reflect the true values of GHGE and BWU of Vietnamese foods. Nevertheless, applying the GHGE value at the food items level minimizes the under- or overestimation compared to only calculating at the food group level. Finally, the environmental impacts of food are not only limited to GHGE and BWU; land use, eutrophication, acidification, and food waste also have environmental consequences. However, limited availability to data on these variables prevented their incorporation into the current study.

To the best of our knowledge, this paper is the first to use national intake data to characterize the differences in environmental impacts from derived dietary patterns—Omnivorous, Traditional and Pescatarian- of Vietnamese women. All dietary patterns have higher but still low nutrient adequacy while, when standardized, showing similar environmental impacts compared to the total population. This indicates that there is need to optimize existing diets for nutrient adequacy and environmental impact. As the distribution of the patterns in the regions and in socio-economic groups differ, strategies to optimize dietary intake should be differentiated between region or income groups. Given the desirability of achieving diets that fulfill the criteria of being both healthy and environmentally sustainable, deliberation on solutions to this issue is warranted.

The datasets presented in this article are not readily available because the dataset is the property of the National Institute of Nutrition (NIN) in Vietnam, and we obtained permission to use it as part of the collaborative effort between NIN and Wageningen University and Research. However, we are not authorized to distribute the dataset to any third party. Requests to access the datasets should be directed to ninvietnam@viendinhduong.vn.

This study was reviewed and approved by the ethics committee of the National Institute of Nutrition, Vietnam. All participants provided informed consent prior to participation in this study.

SN, ET, and IB developed the idea of the manuscript. SN developed the analysis plan of the study, conducted the statistical analysis, and wrote the manuscript. SB, ET, IB, EF, and TL contributed to developing the methods, interpreting the results, and improving the writing of the manuscript. All authors reviewed and approved the final version submitted for publication.

This work was undertaken as part of the CGIAR Research Program on Agriculture for Nutrition and Health (A4NH; https://a4nh.cgiar.org/our-research/flagship-1/), and the CGIAR Initiative Sustainable Healthy Diets through Food system Transformation (SHiFT) (Sustainable Healthy Diets - CGIAR; https://www.cgiar.org/initiative/sustainable-healthy-diets/).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.