AbstractCarbohydrate quality is an aetiological factor of diet-related disease. Indices of carbohydrate quality featuring various ratios of carbohydrates-to-dietary fibre-to-sugar have been associated with improved product and/or diet quality in westernised countries. Carbohydrate intake is especially high in Asia Pacific. Thus, this study evaluated the ability of such carbohydrate metrics to discriminate the nutritional quality of carbohydrate-rich packaged foods and beverages in Australia, Malaysia, Singapore, Thailand and the Philippines, with an additional focus on beverages. This evaluation was conducted by comparing product nutritional composition and assessing products against three national nutrient profiling models. Results showed that Australia had the highest proportion of products meeting all metrics, compared to the Southeast Asian countries. Beverages had a low adherence to all metrics compared to solid foods. Across the five countries, both processed food and beverages meeting the metrics generally contained higher dietary fibre, protein, and certain vitamins and minerals whilst having lower energy, total sugars, free sugars, trans fat and cholesterol content compared to products not meeting the metrics. The metrics were also aligned with national nutrient profiling models to identify nutritious products. In conclusion, these metrics allowed us to discriminate product nutritional quality in the countries assessed and are applicable to beverages.
Keywords:carbohydrate quality; nutritional quality; nutrient profiling model; free sugars; dietary fibre; packaged foods; Asia Pacific
IntroductionCarbohydrates are the primary energy source of the human diet. Its intake is especially high in Asian countries, where the average consumption ranges from 65% to >80% of daily energy intake, contributed primarily by starch-based foods (e.g., rice-based foods) [1–3]. In contrast, western countries, such as Australia, have a lower carbohydrate consumption at an average of <55% to 65% of daily energy intake, and consists of a mix of starch-based foods (e.g., breads) and sugars [1,3–5].
Recent scientific literature has established that carbohydrate quality over quantity is more predictive of disease risk, and the topic of how to define carbohydrate quality has raised much debate over the past years in an attempt to unify recommendations beyond single nutrient targets, such as “high in fibre” or “low in sugar” [6–9]. Among the most well-known measures, the glycaemic index (GI) and whole grain content have been proposed as markers of carbohydrate quality in a product. While GI is a notable measure of dietary carbohydrate quality [10–12], its use and understanding by the broader population is limited, owing to the measure being too technical and its need to be tested clinically . Whole grain, in contrast, is more broadly known and recognised [14–16], but the lack of universal definition of what constitutes a “whole grain food” and a minimum nutritional requirement have led to heterogeneous quality of products bearing a whole grain claim [17,18]. Thus, there is a need to develop a measure of nutritional quality for carbohydrate-rich products, which is nutritionally credible yet straightforward to communicate and compute.
As there is a general consensus that dietary fibre intake, in particular from whole grain sources, should be promoted [6,7,19,20], and free sugars intake be limited [21,22], novel alternative metrics of carbohydrate quality beyond absolute recommendations of these nutrients have emerged. These are expressed as ratios of total carbohydrates and/or starch to dietary fibre. AlEssa and team reported that diets with low total carbohydrates-to-cereal fibre ratios are associated with lower risk for type 2 diabetes mellitus (T2DM) and coronary heart disease in the United States of America (USA) [6,19]. Recently, Blumfield et al., 2020, showed that diets compliant with similar carbohydrate metrics but integrating a free sugar threshold had an improved overall nutritional intake in the Australian population, reflected by a higher Healthy Eating Index score .
While the relevance of such metrics has been demonstrated at diet level, it remains a challenge for consumers to relate their daily food selections with absolute nutrient recommendations. Packaged foods and beverages make up a large part of the modern-day urban diet, and the ability to select healthful products is a key contribution to healthier diets [24–26]. Recently, such metrics have also been proven useful at a product level, with products meeting these metrics found to be nutritionally superior in the USA [17,27] and Brazil . Though the metric has been established in these countries, di erences in eating culture and food regulation have a vast influence on the availability and choice of products. For instance, there is a greater history and more extensive range of plant-based milk substitutes in Asia [29,30], and cereal-based drinks, which are consumed extensively as a snack and/or breakfast, are non-negligible sources of carbohydrate intake in these countries . However, due to the low availability of such products in the USA and Brazil, previous studies [27,28] failed to capture whether the carbohydrate metrics could be relevant to beverages. Nutrient labelling and permitted product claims also influence product formulations in the market significantly. For example, unlike the USA, labelling of added and/or free sugars is voluntary in the countries assessed in this paper, and consumer knowledge of its importance may be limited. Thus, whether the metrics developed based on USA-style dietary patterns may apply to products in Asia Pacific, and especially beverages, remains unknown.
The objective of this study was to investigate whether these carbohydrate metrics could help identify products, particularly beverages, of a higher nutritional quality in countries from the Asia Pacific region, namely Australia, Malaysia, Singapore, Thailand and the Philippines.
Materials and Methods
2.1 Definition of the Carbohydrate Metrics
Three carbohydrate metrics were assessed; these are, per 10 g of total carbohydrates in a product:
1. At least 1 g of dietary fibre (simple ratio);
2. At least 1 g of dietary fibre and no more than 2 g of free sugars (modified ratio);
3. At least 1 g of dietary fibre, and no more than 2 g of free sugar per 1 g of dietary fibre (dual ratio).
The simple ratio was developed by the American Heart Association (AHA) and follows a recommendation based on the ratio of total carbohydrates to dietary fibre in whole wheat . The modified ratio includes an upper limit for free sugars based on the WHO free sugars recommendations described above , and on an average recommendation that about 50% of total energy intake is derived from carbohydrates, then individuals should consume no more than 2 g of free sugar per 10 g of carbohydrates. The dual ratio was developed to put emphasis on the dietary fibre content, rather than the total carbohydrates. In this metric, free sugar content is restricted to 2 g for every 1 g of dietary fibre in the product . This is based on AHA recommendations to consume at least 25 g of dietary fibre  and the WHO free sugar recommendation (equivalent to 50 g of free sugars for a diet of 2000 kcal a day). The dual and modified ratio are subsets of the simple ratio and take free sugars into account.
2.2 Food Databases and Product Selection Criteria
Carbohydrate-based packaged foods and beverages from two di erent databases—Australian Food Composition Database Release 1 (AFCD-1) and Mintel Global New Products Database (Mintel Database)—were curated. Data from AFCD-1 are available as food and beverage sub-categories, whereas data from the Mintel databases are of individual products. AFCD-1 is a national database that contains the nutrient composition of common Australian food and beverage sub-categories . Data were obtained predominantly from analysis of typical Australian products within a sub-category, and the remainder were derived from product labels, imputations or burrowed from other countries. The Mintel Database is a collation of products launched in a country, with details of its on-pack product information including its declared nutrient content, product description, ingredients list and claims . To ensure identification of the most relevant products—predominantly cereal-based—packaged foods and beverages with more than 50% energy from carbohydrates were included in the analysis.
A total of 127 carbohydrate-based packaged food and beverage sub-categories in AFCD-1 were assessed. Several sub-categories in the database were similar in sampling and nutritional composition (e.g., “coconut, fresh, mature, water or juice” and “coconut, fresh, young or immature, water or juice”) and were merged, resulting in a revised total of 101 sub-categories, including 86 food sub-categories and 15 beverage sub-categories.
A total of 8390 carbohydrate-based packaged food and beverage products in the Mintel Database were assessed. These included products from Australia (Mintel Australia) and four Southeast Asian countries (Mintel Asia) during the period of January 2014–August 2019. Mintel Asia consisted of products from Malaysia, Singapore, Thailand and the Philippines.
All sub-categories from AFCD-1 and products from the Mintel Database assessed were classified into 12 food categories and 6 beverage categories (Table 1). Analyses on beverages were separated into ready-to-drink (RTD) and powdered beverages due to differences in nutrient declaration between the two categories (i.e., “per 100 g” for powdered beverages and “per 100 mL” for RTD beverages).
2.3 Proportion of Carbohydrate-Based Packaged Foods and Beverages Based on the Carbohydrate Metrics
The sub-categories from AFCD-1 were weighted by the number of products available on the market for each category using the Mintel Database and then assessed for whether they passed each of the three carbohydrate metrics.
The products from the Mintel Database were evaluated for their ability to meet each metric using their declared carbohydrates, dietary fibre and sugar content. As it is not mandatory for free sugars to be declared on packaging in these countries, the former was obtained through an imputation from the product’s total sugars content using a modified methodology adopted from Louie et al., 2015  (Figure 1).
A product category was considered to have a high adherence to a metric if at least 40% of products in the category met its criteria (highest quartile of product categories), and moderate adherence to the metric if at least 20% but less than 40% of products in the category met its criteria. A product category was considered to have a low adherence to a metric if less than 20% of products met the metric.
2.4. Nutritional Quality of Carbohydrate-Based Packaged Foods and Beverages Based on the Carbohydrate Metrics
2.4.1. Evaluation of Product Nutritional Composition
The nutritional quality between foods and beverages that passed and failed each metric was first evaluated by comparing the level of nutrients. For AFCD-1, this was carried out by using the energy, protein, total fat, saturated fat, trans fat, cholesterol, total sugar, free sugar, dietary fiber, sodium, calcium, potassium, iron, iodine, magnesium, zinc, selenium and vitamins A (retinol equivalent), E, B1, B2, B3, B6, B9 and B12 values of sub-categories from the database.
For the Mintel Database, the nutritional comparison between products that passed or failed each metric was assessed by using the declared energy, carbohydrates, protein, total fat, saturated fat,total sugar, dietary fiber and sodium content. An additional category-specific analysis was performed on categories that had the highest adherence to the metrics—hot cereals, cold cereals, cereal and fruit bars and breads (unfilled)
Due to large variations in serving size across the range of sub-categories and products assessed,all nutrient comparisons were conducted as per 100 g/100 mL of product.
2.4.2. Evaluation of Products against Nutrient Profilin Models In addition to comparing individual nutrients, products from the Mintel Database were also evaluated against national nutrient profilin models specific to the region. Products from Mintel Australia were evaluated against the Food Standards Australia New Zealand Nutrient Profilin Score Criterion (NPSC) and the Australia New Zealand Health Star Rating (HSR). Products from Mintel Asia were assessed against NPSC, HSR and the Singapore Healthier Choice Symbol (HCS). Under the guidelines of the respective nutrient profilin models, a product is considered a nutritious product if it is within the category-specific score limit for the NPSC, 3.5 stars for the HSR or passed the HCS [37–39].The fruits, vegetable, nut and legume content and whole grain content of products were required for this assessment; where undeclared, these values were imputed to complete this assessment.
This evaluation was not carried out for the AFCD-1 as the database is comprised of the nutrient composition of sub-categories and not individual products.
2.5. Proportion of Whole Grain Product Choices and Their Association with the Carbohydrate Metrics
As each carbohydrate metric stemmed from the carbohydrates and dietary fibre content of whole grains, the number of whole grain product choices from Mintel Database (i.e., products that had wholegrain claims or communication) were quantified. To determine if this was associated with the metrics, the parameter was compared with products that passed and failed each metric.
2.6. Comparison of Data from AFCD-1 Mintel Australia
A qualitative comparison of product categories that met each metric from the two Australian databases (AFCD-1 and Mintel Australia) was carried out.
The imputed free sugars content used in the Mintel Database was compared against the free sugars content of sub-categories from the same product category in AFCD-1.
2.7. Statistical Analysis Unequal variances t-tests were performed to analyse differences in nutritional composition between product sub-categories from AFCD-1 and products from the Mintel Database that passed and failed the metrics.
The Pearson’s chi-square test was conducted to determine significant associations between the carbohydrate metrics and whole grain product variants amongst products from the Mintel Database.
To validate that the free sugars estimation calculated for Mintel database using the modified Louie methodology was similar to that of AFCD-1, a paired two-tailed t-test was performed and the Nash–Sutcli e e ciency (NSE) model coe cient was determined.
All statistical analyses were performed using Mintab 18 and Microsoft Excel 2016, and p < 0.05 was used as the criterion to determine statistical significance.
3.1. Proportion of Carbohydrate-Based Packaged Foods and Beverages Based on the Carbohydrate Metrics
In the AFCD-1, all food sub-categories that passed the simple ratio (32%) also passed the dual and modified ratios, whilst only one beverage sub-category passed the simple ratio (2%) and no beverage sub-categories passed the dual or modified ratio (Figure 2a).