Describing a new food group classification system for UK biobank: analysis of food groups and sources of macro- and micronutrients in 208,200 participants

The UK Biobank study is a national prospective cohort in the UK involving 502,655 participants aged 40–69 years at baseline who were recruited between 2006 and 2010 [7]. Approximately 9.2 million eligible adults living within 25 miles of the UK Biobank assessment centres (England, Wales and Scotland) were invited by letter through NHS central registries [8]. Participants who volunteered to take part (5.5% response rate) completed a full baseline assessment with self-reported measurements via touch-screen questionnaires as well as a verbal interview. A wide range of information on socio-demographic factors, lifestyle, and behavioural factors including a short food-frequency questionnaire and medical history were collected, along with physical measurements (such as height and weight), blood and urine samples.

UK Biobank protocols and study details can be found elsewhere (http://​www.​ukbiobank.​ac.​uk/​wp-content/​uploads/​2011/​11/​UK-Biobank-Protocol.​pdf) [7]. The UK Biobank study was conducted according to the Declaration of Helsinki and ethical approval was granted by the North West Multi-Centre Research Ethics Committee (reference number 06/MRE08/65). At recruitment, all participants gave informed consent to participate and be followed-up through data-linkage.

Towards the end of the baseline assessment period (April 2009-September 2010), the UK Biobank started collecting detailed dietary intake measures using a web-based self-administered 24 h dietary assessment, the Oxford WebQ. The Oxford WebQ was completed by 70,724 participants attending their baseline assessment. Additionally, between 2011 and 2012 all participants with valid email addresses (n = 331,013) were invited to complete the Oxford WebQ on four separate occasions. This was done every 3–4 months on variable days to maximise the coverage by season and day of the week (cycle 1: February 2011 to April 2011; cycle 2: June 2011 to September 2011; cycle 3: October 2011 to December 2011; cycle 4: April 2012 to June 2012). Approximately 53% of the participants (n = 176,012) who were contacted by email completed at least one assessment, with a total of 211,050 participants completing at least one dietary assessment either online or at the baseline assessment [6].

The Oxford WebQ collects information on foods and beverages consumed over the previous day. Participants were presented with a list of up to 206 foods and 32 beverages commonly consumed in the UK and selected the number of portions consumed from each food. This food list was constructed using information from the UK National Diet and Nutrition Survey (NDNS) as well as a pilot study [5]. At the start of the questionnaire, participants indicated whether their diet over the previous day was typical and if they were following a special diet. Descriptions and help sections were used to help estimating portion sizes of foods (e.g. slices, cups, servings) and participants were asked to report ingredients of composite dishes separately. Total energy and nutrient intakes were generated by multiplying the number of portions consumed by the set quantity of each food portion size and its nutrient composition obtained from the UK Nutrient Databank Food Composition Tables (FCT) from survey year 6, (2012–2013 and 2013–2014) [9‐11]. Dietary fibre was calculated using the Englyst method [12], which includes non-starch polysaccharides but not lignin and resistant starches.

We classified the 206 foods and 32 beverages reported in the Oxford WebQ into 93 groups (79 food and 14 beverage groups) belonging to 15 main food categories (13 food and 2 beverage categories, Supplemental Table S1). This food group system was mainly based on the classification used in the UK NDNS but many food groups were further disaggregated to offer the potential to investigate a variety of specific research questions which are related to nutrients which may be differently related to health when consumed as part of different foods; such as free sugars, saturated fat or fibre, as well as culinary use; and the extent of processing or plant/animal origin. For cereal and cereal products, we separated breads and pasta by fibre content (e.g. white and wholemeal), while breakfast cereals were divided by type of cereal (e.g. oat and wheat based) and sugar content. Mixed dishes were divided into pizza, other cereal-based dishes with added fat, Indian snacks, sushi and soups, recognising the differences in fat content of each dish as well as their different cultural and culinary roles. For dairy and dairy products, we separated milk, cheese and yogurt by fat content (e.g. higher and medium/lower fat). Fats, butter and spreads were separated by fat content but also by the source of fat (e.g. animal and vegetable fat). Meat and meat products were divided by the type of animal (e.g. beef and pork) except for the group processed meat which may include more than one animal source. Fish and fish dishes were separated into white fish/tinned tuna, oily fish and battered fish. Meat substitutes were divided into soy-based and other vegetarian meals. Vegetables were separated into groups which considered micronutrient, carbohydrate or protein content (e.g. green leafy/cabbage, root (excluding potatoes), tomatoes, allium, legumes) as well as fat content (e.g. baked/boiled potatoes; mashed potatoes and fried/roast potatoes). Fruits were grouped according to micronutrient content (e.g. citrus; berries and apples/pears), as well as processing (e.g. dried and stewed fruit). Nuts and seeds were divided according to the salt content. Sugary foods were divided into groups reflecting differences in consumption: added sugars/preserves (including table sugar); chocolate confectionery; other sweets (non-chocolate); biscuits/cookies; desserts, cakes and pastries; milk-based desserts; soy-based desserts; and sweet spreads (including chocolate and peanut-butter spreads). Sauces and condiments were divided by the fat content. Beverages were first separated into alcoholic and non-alcoholic. Non-alcoholic beverages were further separated by the caffeine content, sugar content, or presence of milk. Where possible, milk added to coffee/tea and porridge was disaggregated to be included in the milk group. For other milk-based drinks this was not possible and they were categorised in their own group. Alcoholic drinks were then divided into wine, beer/cider, and spirits.

The final food group classification was further refined and consolidated following consultation with three nutrition scientists from the UK with experience in nutritional epidemiology or public health policy.

Participants who completed a minimum of one valid WebQs were included. Participants with implausible energy intakes (over- and under-reporters) were excluded before analysis. We used the individualised method to calculate the ratio of reported energy intake (EI) to estimated energy requirement (EER) (EI:EER), where EERs were calculated using the Schofield Equation [13]. 95% CIs were calculated to classify individuals as plausible reporters (EI:EER within the 95% CIs), over-reporters (EI:EER > upper 95% CI) or under-reporters (EI:EER < lower 95% CI).

Demographic and lifestyle characteristics were collected at baseline using a touchscreen questionnaire and were categorised as: White ethnicity vs other (including Asian, Black, mixed background); education was classified as higher degree (college or university degree, or professional qualifications) vs any school degree (A levels, AS levels, O levels, GCSEs or CSEs) vs vocational qualifications (NVQ, HND or HNC), vs no qualifications or other not classified elsewhere; physical activity was categorised as high (≥ 3000 metabolic equivalent (MET)-minutes per week) vs moderate (≥ 600 and < 3000 metabolic equivalent (MET)-minutes per week) vs low (< 600 metabolic equivalent (MET)-minutes per week); smoking was categorised as never, current, previous; alcohol intake was categorised as 5 + units/week, 1–4 units/week, < 1 units /week, never. Body mass index (BMI) was calculated using the measured height and weight and categorised as: underweight [< 18.5 kg/m²), healthy weight [18.5 to < 25 kg/m²], overweight [25 to < 30 kg/m²], obesity [> 30 kg/m²].

We calculated the individual’s mean intakes from all their completed dietary assessments. Descriptive statistics (crude means, SD) were computed for total daily energy (kJ/day), weight or volume of food and beverages (g or ml/day), as well as for macronutrients (g/day) and micronutrients (mg or µg/day). For each food and beverage group, we calculated per capita intakes as well as mean intakes among consumers only, including only the participants consuming more than 0 g or kJ from each food group. To identify the major sources of energy intake, the per capita contribution of each food or beverage group to total daily intake was calculated as the energy consumed from each food group divided by the total daily energy. We finally investigated the top food group sources of energy, saturated fat, and free sugars by sex and education status. Stata version 14 (StataCorp LP) was used for all analysis.

The ten major contributors to EI per capita (in descending order) were (Table 1): desserts, cakes, and pastries (5.2% EI); white bread (3.7% EI); white pasta & rice (3.1% EI); bananas & other fruits (other than citrus, berries, apples and pears, 3.1%EI); semi-skimmed milk (3% EI); biscuits (2.9% EI), fried/roast potatoes (2.9% EI); wholemeal bread (2.7% EI) and mixed bread (2.7% EI); and high-fat cheese (2.7% EI). Among the beverage subcategories, the top contributors to energy intake were fruit juice (1.8% EI) among non-alcoholic beverages and red wine (2.4% EI) and beer/cider (2.4% EI) among alcoholic beverages. However, water/sparkling water (513 g/d) and tea (425 g/d) were the top contributors to beverage volume per day. Overall, the categories of cereal and cereal products, followed by the sugar, preserves, cakes & confectionery and vegetables and potatoes were the top 3 categories contributing the most to the total energy intakes (Table 1).

Figure 1 shows the top food contributors to relevant macronutrients. Desserts, cakes and pastries were top contributors to total carbohydrate, free sugar, total fat and saturated fat intakes (Fig. 1, Supplemental Table S3). The bread was the top contributor to total carbohydrate (white, mixed) while wholemeal and mixed bread was the top contributor to fibre intake. Bananas & other fruits were major contributors to both total carbohydrates and fibre intakes. High fat cheese was a top contributor to total fat, saturated fat, and protein intake. Dairy fat spreads (e.g. butter) were top contributors to both total fat and saturated fat. Poultry and beef were the top sources of total protein intakes. Semi-skimmed milk was an important contributor to many micronutrients, including vitamin B12, calcium and potassium intakes (Supplemental Table S4). Overall, fruits and vegetables, as well as fruit juice made a substantial contribution to intakes of folate, vitamin C, and potassium. Beef, oily fish were also important contributors to vitamin B12, and beef alongside bread and cereals were important contributors to intakes of iron.

Total daily energy and alcohol intake were higher in men than in women, but the proportion of energy from macronutrients was similar (Table 2). The top ten foods contributing the most to energy intake were similar in men and women (Fig. 2), with desserts, cakes and pastries ranking at the top. Among men, white bread and beer/cider ranked 2nd and 3rd; whereas in women bananas and other fruit and white pasta/rice ranked 2nd and 3rd. Fruit juice was the top contributor to free sugars in women, while added sugars/preserves was the top contributor among men. For SFA intake, high fat cheese, desserts and cakes, and dairy-fat spreads provided the most SFA to the diets among men and women.

Total daily energy and macronutrient intakes were similar regardless of education (Table 2). Desserts, cakes and pastries and white bread were the top sources of energy intake in both education groups (Fig. 2). A few small differences were noted. Among people with higher education, white pasta/rice and higher fat cheese provided a higher proportion of energy intake, while those with lower education obtained a higher proportion of energy from biscuits, white bread, fried/roast potatoes and beer/cider. Fruit juice contributed more free sugars among people with higher education, whereas added sugars and preserves, sugars-sweetened beverages (SSB) and beer & cider provided more free sugars among those with lower education. Differences in SFA sources between education groups were usually small (< 0.6 g), but high fat cheese, dairy fats, desserts & cakes made a greater contribution to saturated fat in the diets of people with higher education, whereas biscuits, chocolate confectionery and beef provided slightly more saturated fat to the diet among those with lower education.