2. Development and Findings
This study used a score based on the food and nutrient recommendations underpinning the EWG and a dietary index based on the NPM versions of 2005 and 2018 to evaluate the healthiness of representative diets of the UK population. It demonstrated that, overall, diets considered healthy according to the EWG score achieve a lower NPM dietary index score, irrespective of the version used to calculate that index. The mean NPM dietary index was higher for the 2018 version than the 2005 version of the NPM at each level of the EWG score. Overall, agreement between the EWG score and the NPM dietary index in classifying a diet as “healthy” was good when a low cut-off for the NPM dietary index was used, suggesting that the NPM is broadly consistent with the UK’s FBDGs.
Both FBDGs and NPMs are based on the principle that foods can be classified as healthy if their consumption is associated with a reduced risk of disease or improved health and wellbeing 
. However, the association between diet and health is complex and observational studies are subject to confounding because individuals who eat “healthy” diets tend to engage with other health-promoting behaviours and lifestyles, live in more affluent areas, and have a higher education level 
. In addition, a diet that reduces the risk of disease, i.e., a healthy diet, depends not only on the individual foods that constitute the diet, but also on the frequency, amount, and combination in which they are eaten 
. Therefore, it is difficult to compare NPMs, which assess the healthiness of individual foods, and FBDGs, which recommend what to eat to have a healthy diet. To add complexity, FBDGs do not, for most foods, classify them as either healthy or unhealthy, but rather make recommendations about the composition of a healthy diet based on broad food groups, and within these food groups there is a wide variability in nutrient composition.
Nonetheless, for some interventions aimed at improving health in the population, it is necessary to categorise foods according to their contribution to the healthiness of the diet. NPMs can serve as the basis for this categorisation, and it is thus important to assess whether this categorisation is consistent with other government advice on healthy diets. This study found that the healthiness of actual diets (as measured by compliance with the recommendations underpinning the EWG) reflects differences in the proportions of healthy and/or unhealthy foods as assessed by the NPM. Overall, it supported this assumption, because as the NPM dietary index decreased as the EWG score increased, irrespective of which version of the NPM was used. In addition, healthier diets, i.e., diets with a higher EWG score, had a narrower distribution of NPM dietary index values than less healthy diets. This suggests that healthier diets, as assessed by the EWG score, tended to include mostly foods that would be classified as healthy by the NPM, whilst less healthy diets included foods within a broader range of healthiness according to the NPM. However, the sample size for high and low EWG scores was small, and thus, the findings need to be interpreted with caution.
Although there is no gold standard against which to compare NPMs, different methods have been used to validate NPMs, such as comparing NPMs against FBDGs or health outcomes 
. There is substantial overlap among FBDGs worldwide, as these are based on evidence on the components of a healthy diet that are associated with a reduced risk of nutrition-related diseases 
. In keeping with this, NPMs should rank foods according to their healthiness, which should similarly reflect a reduced risk of nutrition-related diseases. Prospective cohort studies have shown an association between adherence to the recommendations underpinning the EWG and improved health outcomes 
. Therefore, showing that the NPM dietary index is broadly concordant with the EWG score suggests that it ranks foods as healthy and less healthy by applying similar criteria to those employed by the EWG, which have been shown to be associated with improved health. Although further studies are required to confirm whether the 2018 version of the NPM is more consistent with the EWG than the 2005 version, there are possible explanations why this might be the case. First, the recommendation regarding fibre was poorly met by those with low EEG scores in general, and the updated version of the NPM increased the daily requirement for fibre from 24 g to 30 g and increased the points afforded to fibre. This meant that high-fibre foods would achieve lower NPM values (as fibre points are deducted), and those foods would be included in greater amounts in diets with higher EWG scores. Second, the more restrictive allowance for free sugars in the 2018 version of the NPM may have increased its alignment with the EWG, as the recommendation related to free sugars was most commonly met by those with high EWG scores.
NPMs have been broadly used for two main purposes: supporting consumers facing food labelling and the regulation of food marketing and advertising. Although the principles and criteria underpinning NPMs were developed for both purposes and are similar, the way in which they are used can be different. Food labelling based on NPMs can assume that foods are distributed along a continuum of relative nutritional quality ranging from healthier to less healthy 
. These food labelling systems, such as those used in Australia or France, are typically graded systems that rank the nutritional quality of foods across the range of possible NPM values 
. The healthiness of foods is then displayed using a score that is depicted as stars, letters, or colours. NPMs can also be applied not as a continuous, but as a binary measure, that either allows or prohibits marketing of certain foods and drinks, as happens in the UK. This means that its alignment with the EWG varies according to the threshold that defines which foods are unhealthy, and hence, which are subject to marketing restrictions. Due to the way the NPM dietary index was calculated in this study, it was not possible to determine directly the exact value of NPM that should be used as cut-off to classify foods as healthy or less healthy in order to maximise concordance with the EWG.
Although the scientific merit of applying a binary definition of individual foods as “healthy” or “unhealthy” based on NPM has been debated 
, even in countries where NPMs are used as continuous scores for food labelling, pressure has been mounting to adopt objective criteria defining “unhealthy” foods in order to regulate marketing and advertising 
. Decisions about whether a certain food can or cannot be advertised require a binary definition of “healthy” and “unhealthy” based on a pre-specified cut-off value of the NPM. However, there is no scientific consensus on the existence of a specific nutritional composition threshold that distinguishes between “healthy” and “unhealthy” foods. For the purpose of regulation and taxation, a binary classification may be unavoidable, but this needs to be carefully explained to the population to avoid unintended consequences of determining that foods are “healthy” or “unhealthy” 
National FBDGs provide the overarching framework and benchmark for a healthy diet, based on current knowledge of the associations between various dietary components and health outcomes 
. It is, thus, important to ensure that the EWG, which provides the official advice on healthy eating to food manufacturers, retailers, and consumers, is consistent with the NPM, which is used to regulate marketing of foods in the country. Otherwise, the population will get mixed messages about healthy diet and food, which can exacerbate the ongoing problem of misleading nutritional claims on food labels and adverts 
. This study demonstrated that, overall, the EWG score and the NPM dietary index agree on what constitutes a healthy diet, for low values of the NPM index, using either the 2005 or 2018 versions. This suggests that there is good alignment between the NPM and the EWG. It is important to note, though, that the overlap between NPMs and FBDGs can never be perfect. For example, while salmon falls into the recommended food group of fatty fish, its high fat and salt content, particularly for smoked salmon, can render it “unhealthy” according to the NPM. Rather than invalidating NPMs or FBDGs, these discrepancies emphasise the complementarity between the two approaches at food and diet levels and highlight the need for clear guidance to the public so that they understand these nuances when making food and diet choices.
3. Strengths and Limitations
This study has several strengths. It used actual diets from individuals living in the UK to compare the EWG score and the NPM dietary index, which are more relevant than hypothetical diets. Actual diets can take greater account of other factors that are unrelated to health that shape diets (such as the palatability of food), rather than modelled diets. In addition, nutritional composition was available for all foods in a standardised format, which enabled computing the NPM and estimating compliance with the recommendations underpinning the EWG with accuracy.
There are also some limitations worth acknowledging. First, the NDNS has a relatively small sample size, which limited the ability to perform a subgroup analysis according to age, sex, or region. Second, the NDNS relies on self-reported food intake, which may be subject to bias. Third, the NPM dietary index was computed as a continuous variable to grade the nutritional quality of individual diets, whilst this particular NPM was designed to be used to classify foods as healthy or less healthy using a cut-off of four for foods and one for drinks. Therefore, the NPM dietary index cannot be directly interpreted or compared with the NPM value that is calculated for individual foods. In addition, although the NPM 2005 and NPM 2018 categorise foods and drinks as “less healthy” using the same threshold, they are not directly comparable as ordinal measures due to changes to the scales used. Fourth, it is possible that agreement between the EWG score and the NPM dietary index varies according to the context and population (e.g., alignment may differ between adults and children). However, those limitations are unlikely to have had a material impact on the key findings of this study. Fifth, salt consumption was based on salt that is included in foods (either naturally occurring or added during processing), but not salt added at the table, which means that individual salt consumption may be underestimated. Sixth, we applied generic food and nutrient recommendations to the entire population instead of sex- and age-specific recommendations. Nonetheless, this had no material impact on the study findings as the purpose was to compare different classification systems, which were applied irrespective of demographic characteristics of the individuals. Age- and sex-specific recommendations would have been important if the aim was to evaluate individual diets.