In assessing diet quality, the need to analyse dietary patterns, rather than single dietary components, is increasingly recognized. This is because people consume combinations of foods, and dietary components may interact.
Two approaches to dietary patterning have been used: theoretically defined dietary patterns and empirically derived dietary patterns. The latter are statistically derived ‘a posteriori’ from collected food consumption data based on clustering of individuals or correlations in intakes of the various dietary components . As empirically derived dietary patterns may be very variable between countries and over time, they are not considered useful for the purpose of INFORMAS. Consequently, theoretically defined dietary patterns (or pre-defined measures of diet quality) were the focus of this review.
Indices of overall diet quality based on selected foods and nutrients
Indices of diet quality are pre-defined summary measures of overall diet quality, and can be used to monitor overall dietary changes . They are based on foods and/or nutrients that are considered important for health in general or for specific health outcomes. In the last decades, several indices have been developed, and some of them have been validated by relating them to measures of overall nutrient adequacy or chronic disease risk. The Healthy Eating Index (HEI) , the Diet Quality Index (DQI) , the Healthy Diet Indicator (HDI)  and the Mediterranean Diet Score (MDS)  are the four original diet quality scores that have been used most extensively. However, the original HEI is no longer used and has been superseded by the HEI-2005  and the HEI-2010 .
The original HEI, based on the 1995 Dietary Guidelines for Americans, comprises grains, vegetables, fruits, milk, meat, total fat (as a % of total energy), saturated fat (as a % of total energy), cholesterol, sodium and dietary variety as components . Scores between 0 and 10 were assigned to each component and an index with values between 0 and 100 obtained. After the 2005 Dietary Guidelines for Americans were released, the index was modified to address the increased importance of whole grains and specific types of fats , and included 12 components: total fruit, whole fruit, total vegetables, dark green and orange vegetables and legumes, total grains, whole grains, milk, meat and beans and oils, saturated fat (as a % of total energy), sodium in grams per 1,000 kcal and the calories from solid fat, alcohol and sugar (as a % of total energy). In addition, the index was developed to assess intake densities rather than absolute intakes, which makes it adaptable to any set of foods at any level of the food system and not just individual level intakes . The components are weighted such that each major food group mentioned in the guidelines receives basically equal weight except empty calories, which are double weighted in accordance with their effect on diets. The most important changes from 2005 to 2010 include the addition of seafood and plant proteins to capture specific choices from the protein group; a ratio of polyunsaturated and monounsaturated to saturated fatty acids, replacing oils and saturated fat; and a moderation component, refined grains, replacing total grains, to assess overconsumption .
The DQI comprises total fat, saturated fat, cholesterol, fruits and vegetables, grains and legumes, protein, sodium and calcium as components . Scores 0, 1 and 2 are assigned to each component, and the index ranges from 0 (excellent diet) to 16 (poor diet). In 1999, this index was revised to incorporate measures of dietary variety and moderation , and consisted of 10 components, with scores between 0 and 10 assigned to each component. Total score ranged from 0 (poor compliance) to 100.
The HDI, based on WHO recommendations, includes saturated fat, polyunsaturated fat, protein, complex carbohydrates, dietary fibre, fruits and vegetables, pulses, nuts and seeds, mono- and disaccharides, and cholesterol as components . Scores 0 and 1 are attributed to each component and an index ranging between 0 and 9 is obtained.
The MDS includes the ratio of monounsaturated fat to saturated fat, legumes, cereals, fruits and nuts, vegetables, meat (products), milk and dietary products, and alcohol as components . Score 0 or 1 is assigned to each component using the median intake of study participants as a sex-specific cut-off point. A score ranging from 0 (low adherence to the Mediterranean diet) to 8 is obtained. In 2003, fish intake was included as an additional component, retrieving scores ranging from 0 to 9 .
Waijers et al.  reviewed the Medline literature published until September 2005 on 20 pre-defined indices of overall diet quality. They found only modest associations between diet quality indices and either nutrient adequacy or health outcomes and concluded that existing indices do not predict mortality or morbidity significantly better than individual dietary factors. Arvaniti et al.  reviewed 23 commonly used dietary indices (overlapping largely with those from Waijers et al. ) in a study published in 2008, and concluded, as did Waijers et al. , that many arbitrary choices are included in the make-up of the scores and that the majority of the indices fail to recognize the different interrelationships between their constituents. They recommend the development of weighted indices, with weights in proportion to the importance of a food/nutrient to the health outcome of interest . A systematic review of published English-language literature until 2007 among adults was conducted by Wirt et al. , including 28 articles and 25 indices of overall diet quality, with HEI , HDI , Healthy Food Index , the Recommended Food Score , the DQI , the Diet Quality Score  and MDS  as the most important. The majority of studies presented methodological weaknesses but demonstrated that higher diet quality was consistently inversely related to all-cause mortality with a protective effect of moderate magnitude. The predictive capacity of most of the indices was fairly similar . Kourlaba et al.  concluded that, based on a Medline review until June 2008, pre-existing indices are adequate tools to evaluate diet quality, but they have shown moderate predictive ability in relation to chronic diseases and health determinants. The predictive capacity of MDS or adapted versions was shown to be slightly better than that of the original HEI, DQI or HDI . A systematic search of Medline for prospective cohort studies or randomized trials investigating dietary exposures in relation to coronary heart diseases found that the MDS is a better predictor of coronary heart disease than individual foods . Two more recent systematic reviews were identified [128, 129], but they only included studies using indices among children and are not further discussed here.
In order to update the existing systematic reviews with more recent literature, Medline was searched for the terms ‘Healthy Eating Index’, ‘Mediterranean Diet Score’, ‘Diet Quality Index’ and ‘Healthy Diet Indicator’ (four most extensively used dietary indices) for the period 2008–2012, and studies linking diet quality with overweight, obesity or different diet-related NCDs were included. In total, 64 studies [130-193] were retrieved using one or more of these indices or their modified versions. The (original) HEI (32 studies) and MDS (38 studies) were most frequently used (some studies used multiple indices).
In general, more positive than null associations were found with both overweight/obesity and NCDs, but predictive ability in relation to NCDs was moderate, such as that found in the prior reviews. Part of the reason for this might be related to measurement error, about which we know very little in the case of multivariate dietary patterns. This is an important area of research that needs to be pursued.
A new indicator of diet quality based on the share of ultra-processed food products in the diet
The dietary share of ultra-processed food products, expressed as a percentage of total calories, has been recently proposed as a predictive indicator of the energy and nutrient adequacy of the overall diet [5, 194]. Ultra-processed products are ready-to-consume industry formulations manufactured from ingredients directly extracted from whole foods, such as oils, fats, sucrose and flours, or processed from components extracted from whole foods, such as high-fructose corn syrup, hydrogenated oils, a variety of starches and the cheap parts or remnants of meat. To these products, several additives are typically added, with little or no content of whole foods. They include pre-sugared breakfast cereals, sweet and savoury snacks, desserts, a variety of ready-to-heat dishes and soft drinks [5, 194]. The production and consumption of ultra-processed food products is rising rapidly, particularly in LMICs, with the effect of eroding food systems and dietary patterns based on minimally processed foods and freshly prepared meals [5, 88, 195-197].
Ultra-processed food products typically are energy dense; have a high glycaemic load; are low in dietary fibre, micronutrients and phytochemicals; and are high in unhealthy types of dietary fat, free sugars and salt . Studies in Brazil and Canada have documented the direct association of the dietary share of ultra-processed products with overall dietary energy density and the content of free sugars, total and saturated fats, and sodium. These studies have also shown an inverse association between the dietary share of ultra-processed products and the overall protein and fibre content of the diet [4, 5]. Preliminary evidence documents direct association of the dietary share of ultra-processed products with the risk of metabolic syndrome , and overweight and obesity . However, it is noted that this new indicator has not been thoroughly tested yet.