During the past decades, a worldwide rapid nutritional transition has been observed, in which large shifts occur in diet as well as in physical activity and lifestyle( Reference Popkin 1 ). The global Westernisation of diet shares common characteristics across countries with an increased intake of animal source foods, added sugar, energy from fats and energy-dense foods( Reference Popkin 2 – Reference Sook 4 ). This nutritional change has been hypothesised to be associated with an increased incidence of and mortality from major non-communicable chronic diseases( Reference Popkin 5 ). The Republic of Korea (South Korea) is a country that has experienced one of the world's fastest socio-economic growth called ‘the Miracle on the Han River’ since the outbreak of the Korean War( Reference Kim, Moon and Popkin 6 ). These rapid socio-economic and socio-cultural changes have remoulded the country's tradition and peoples' lifestyle affecting the unique aspects of the traditional diets in Korea( Reference Lee, Popkin and Kim 7 ). Recently, a new type of dietary pattern has been identified in addition to Western and traditional diets( Reference Lim, Kim and Wang 8 ). The trend of nutrition transition is uneven between countries – probably reflecting different phases in this dynamic process and country-specific diets and environment. For example, the proportion of fat-derived energy in Korea is low compared with its neighbouring countries (China and Japan) even though the total intake of fat is gradually increasing in Korea compared with other Asian countries( Reference Kim, Moon and Popkin 6 , Reference Lee, Popkin and Kim 7 , Reference Zhai, Wang and Du 9 ). When it comes to obesity patterns, Malaysia and Japan tend to have a higher prevalence of obesity while, in comparison, China and The Philippines have a relatively lower prevalence of obesity( Reference Kim, Moon and Popkin 6 ). This country-specific nutritional transition and its association with cancer and other non-communicable diseases (NCD) need to be better understood to reverse and prevent the dramatically increasing trends on diet-related NCD burden projected in countries undergoing nutritional transition worldwide.
Applying a standardised methodology will improve the quality and comparability of data on dietary exposure assessment as diet is one of the main factors affecting the current NCD burden across countries. Indeed, the problem faced in tracing the changes in diet is that there are only few studies on diet and nutrition transition using comparable and individual data across countries due to the lack or heterogeneity of the dietary assessment methodologies( Reference Popkin 10 ). This limitation has been found to reduce the possibilities to harmonise observations, recommendations and concerted actions of relevance for dietary research, nutritional surveillance and public health, at the international level( Reference Elmadfa, Meyer and Nowak 11 ).
The GloboDiet software (previously named EPIC-Soft), a computerised interview-based 24 h dietary recall program developed at the International Agency for Research on Cancer (IARC), is a highly standardised methodology successfully developed, validated and implemented as a reference dietary methodology in different international epidemiological and surveillance settings( Reference Crispim, Geelen and de Vries 12 – Reference Ocke, Slimani and Brants 17 ). This method was initially used as a reference calibration method between twenty-three centres in the ten countries participating in the European Prospective Investigation into Cancer and Nutrition (EPIC) study, coordinated by the IARC. In the EPIC calibration study, a single 24 h recall was collected by direct or by telephone interview from almost 37 000 participants using an earlier version of GloboDiet( Reference Slimani, Ferrari and Ocke 18 ). More recently, this international software has been adapted, validated and implemented in countries that participated in the pan-European surveillance system( Reference Slimani, Casagrande and Nicolas 19 ).
Benefiting from these successful experiences in Europe, application of the GloboDiet methodology to other geographical regions could contribute to provide more comparable dietary exposure data worldwide, particularly within the framework of the WHO action plans to fight against and prevent the current diet-related NCD burden worldwide. The main objective of the present study was thus to develop a Korean version of the GloboDiet software and its complementary tools, as a first attempt to expand this IARC-WHO international computerised 24 h dietary recall program to an Asian context.
Experimental methods
The concepts of the standardised 24 h dietary recall computerised software (GloboDiet)
The interactive 24 h recall interview software was designed following pre-defined and standardised procedures described in detail elsewhere( Reference Slimani, Deharveng and Charrondiere 20 ). The 24 h recall interview using this software has five main steps: (1) general non-dietary information; (2) quick list; (3) description and quantification of foods and recipes; (4) quality controls; (5) description and quantification of dietary supplements. The structure of the software and detailed information of interview steps have been described in detail elsewhere( Reference Slimani, Casagrande and Nicolas 19 ).
Overall structure of databases
About seventy databases on foods, recipes and dietary supplements were customised and translated into Korean at the IARC. Common databases are the backbone of the standardisation within and between countries, for instance food and recipe classifications, facets/descriptors to describe foods in a comparable way, quantification methods and probing questions. ‘Facets’ are series of systematic questions to code different characteristics of the food (e.g. cooking method and preservation method) while ‘descriptors’ are pre-defined potential answers associated with each facet (e.g. boiled, fried and steamed for the ‘cooking method’ facet). More detailed explanations about the concepts and definitions of ‘facets/descriptors’ have been published elsewhere( Reference Slimani, Ferrari and Ocke 18 , Reference Slimani, Deharveng and Charrondiere 20 ). In contrast, country-specific databases are meant to capture the differences in diet existing within and between countries, for instance, food and recipe lists; brand name lists; facets and probing, which are specific to foods and recipes; quantification methods for each food item and recipe; coefficients for edible portion; cooking; density; fat left on the dish; default ratios for fat, sauce and sweeteners added to foods and recipes; and databases for quality controls( Reference Crispim, Nicolas and Casagrande 21 ). All databases were prepared and customised using the data available in Korea( 22 – 29 ).
Preparation of common databases
Food and recipe classifications were adapted according to foods consumed in Korea based on the IARC database library, which contains the common reference databases used in all GloboDiet versions. Facets/descriptors and probing questions were also selected from the IARC database library and several descriptors were added. For dietary supplements, facets and descriptors were also selected from the database library. All the common files were translated into Korean at the IARC.
Preparation of country-specific databases
Based on the nutrient database and recipes available from the Korean Nutrition Society( 26 ), food and recipe lists were prepared considering their frequency of consumption at the National Cancer Center of Korea (NCC) with the guidance of IARC. Databases of the UK (demo) version were used as a model for database customisation process. Korean foods and recipes were classified according to adapted GloboDiet classification. For fish and vegetables, scientific names were added into their English names to ease their identification. Recipes from the Korean Nutrition Society were loaded into the Recipe Manager application, a specialised module of the GloboDiet methodology to manage recipes, at both the NCC and the IARC. Food items and recipes can be quantified by means of photographs, volumes (directly or as household measures or as shapes and thickness), standard units (e.g. an apple, a can) or weight. Local standard units, shapes and household measures were gathered by the NCC in Korea. The information on the commercial products was collected by the NCC team by surveying at the major markets in Korea and searching on the Internet. Afterwards, the quantification methods such as household measures, photographs and standard units were assigned to each food item and dish, according to groups of foods/dishes (e.g. vegetables and dairy food) and states in which they are consumed, at the NCC and the IARC. During the interview, fractions and multiple of proposed quantity can also be applied to record the actual consumed portion. Coefficient databases including information on density, edible portion, weight change during cooking, absorbed fat weight during cooking, standard percentages of fat, sauce, sugar added to food/recipe and standard percentages of fat left in the dish were customised at the IARC using Korean values collected by the NCC and other values available from the other (European) versions, following the Standard Operating Procedures and guidelines prepared by the IARC. Facets, quantification methods and probing questions were assigned to each food and recipe at the IARC. Nutrient values for energy, carbohydrate, protein, fat and alcohol were attributed to each food item at the NCC using the nutrient database from the Korean Nutrition Society( 26 ) for a final quality control at the end of the interview. Maximum portion sizes for each food item and recipe were set as warning to detect extreme or implausible portion sizes at the IARC. The dietary supplement list that was developed by the Seoul National University of Korea( Reference Kang, Kim and Baek 30 ) using the 4th Korean National Health and Nutrition Examination Survey (KNHANES) contains the most up-to-date dietary supplement data and was adopted as such. All dietary supplements that were consumed more than once by respondents were included in this list. Additionally, texts for the user interface of the software were translated into Korean at the IARC. General information for dietary survey such as places of consumption, date, special occasions and special types of diets were translated into Korean at the IARC as well. We used weight and height data from Korea Centres for Disease Control and Prevention( 23 ) and Korean Agency for Technology and Standards( 28 , 29 ) to adapt the anthropometric database that is used to check missed fields or out-of-range values during subject information data entry; subject information is used for ultimate quality controls on energy intake v. energy requirement calculated according to its sex, age, height and weight( Reference Crispim, Nicolas and Casagrande 21 ).
Picture book
The picture book( 31 ), part of the GloboDiet methodology, contains photographs of foods/dishes, household measures and shapes to assist respondents in estimating portion sizes when using GloboDiet. Photographs of foods/dishes: an ad hoc picture book was made including new Korean foods/dishes that were selected at the IARC as described below. The IARC guidelines( Reference Crispim, Nicolas and Knaze 32 ) were followed at the IARC to choose the portion sizes and by a professional food photographer in Korea to take the pictures at the NCC. For spreads on bread, such as jam and butter, pictures for spreads of the GloboDiet picture book were enclosed in the Korean picture book. In order to select foods/dishes to be included in the picture book, foods and dishes consumed in the fifth KNHANES data( 25 ) were listed by descending order of consumption frequency until cumulated percentages of the consumption frequency reached 80 %. Foods and dishes that can be easily quantified with household measures or shapes were removed from the list, whereas several foods or dishes that were difficult to quantify without pictures were included in the list even though they were not selected in the previous analysis (e.g. somen, Chinese noodles, cellophane noodles and sausage). Portion sizes to get four to six portions per photograph series of each selected food item and dish (see Fig. 2 for photograph series 412) were determined by analysing the intake amounts of the fifth KNHANES data. For the selected foods and dishes, individual portions consumed in one meal were calculated and examined. In principle, 10th and 90th percentile of consumed portions were selected as minimum and maximum portion sizes of pictures, respectively. For several food items, the selected minimum portions were too small and the 25th percentiles of consumed portions were instead selected as minimum portions. On the contrary, if the selected portions corresponding to the 90th percentile were too small, consumed portions of the 95th percentile were selected as maximum portions. These exceptional selections were made to enable each portion to be distinguishable and visible in the picture. From four to six portions were determined with a linear increase of at least 25 %. Shapes: new shapes of foods and dishes were designed based on collected data of packaging sizes surveyed by the NCC at the major market places in Korea with the guidance of IARC( Reference Crispim, Nicolas and Knaze 32 ). Food items and dishes were purchased and sliced if needed. Their surface areas were drawn to be included in the picture book and measured to be used for the quantification. Household measures: pictures of tableware that is generally used at restaurants and home were taken at the NCC according to the IARC protocol( Reference Crispim, Nicolas and Knaze 32 ) (see Fig. 3).
Results
This project provided the first Asian version of GloboDiet, with about seventy databases being customised and translated into Korean. Customisation process of databases is shown in Table 1. New (sub-) subgroups were created and added into the existing common classification in order to fit to Asian specific foods. For food classification, seaweeds were added into ‘vegetables’, and insects and jellied food were added into ‘miscellaneous’ food groups. For the recipe classification, ‘based on eggs’ was added into ‘soups.’ The aim of having (sub-) subgroups in classification is to ease the classification of foods/recipes and comparisons across countries. Indeed, in order to reflect the high consumption of seaweeds in Korea/Asia compared with Europe, seaweeds previously classified under the ‘leafy vegetables’ subgroup were gathered into a new and specific sub-subgroup ‘seaweeds’ created under the ‘leafy vegetables’ subgroup. In the Korean food list, eighteen kinds of seaweeds were reported in contrast to the one to three (generic) items (seaweeds) that were indicated in most of the European food lists. For the description of Korean food items, eight descriptors of seafood such as eel, croaker and squid were added into the facet ‘source’. In addition, twelve new flavours were added into the facet ‘flavoured/added component’ to describe specific Korean foods such as traditional Korean cookies and soya milk. For example, ‘seaweed’ and ‘pine pollen’ were added for traditional Korean cookies while ‘pumpkin’ and ‘sweet potato’ were added for soya milk. Asian fruit flavours such as ‘Japanese apricot/Chinese plum’, ‘wild grape’ and ‘persimmon’ were added for yogurt and drinks. These minor amendments of the common files did not affect the overall structure or general concept of standardisation across country-versions of the GloboDiet software. Country-specific databases were prepared starting from the available Korean databases( 22 , 24 – 27 ) and borrowing European data when necessary. Korean food, recipe, dietary supplement, synonym and brand name lists were collected in Korea. Total numbers of foods, recipes and dietary supplements were 1305, 119 and 1157, respectively. Facets were assigned to each food item and recipe. Databases on quantification methods were adapted to Korean diet considering types of food and Korean food packages available in the market.
Table 1 Customisation of databases for the Korean version
KNHANES, Korean National Health and Nutrition Examination Survey.
A Korean picture book was generated including seventy-four Korean foods/dishes, four pictures for spreads, twenty-five Korean household measures and five shapes. Only four pictures from the original ‘GloboDiet picture book’ were included. The content of the picture book is described in the Table 2. The pictures of foods and dishes comprise two categories: (1) general pictures for quantification of the amount of foods/dishes consumed by respondents and (2) help pictures for identification of the food form/parts linked to standard units.
Table 2 The list of pictures in the Korean picture book
* General pictures enable to quantify the amount of foods consumed by respondents.
† Help pictures help to identify the food size/part and are linked to standard units.
Fig. 1 shows screen shots illustrating the description of a steamed potato by a series of facets and descriptors in the Korean version and UK version as a means of comparison. In each window automatically displayed on the screen, a series of prompts asks the characteristic (facet) of the consumed food such as cooking method, physical state/form as quantified and preservation method. These series of prompts are proposed predefined descriptors (facet-related answers) that were adapted to the Korean food and recipe groups.
Fig. 1 Description of food by a series of facets and descriptors in the Korean and UK versions of GloboDiet: an example of steamed potato (the circled item). A colour version of this figure can be found online at http://www.journals.cambridge.org/bjn
An example of photographs to quantify cabbage kimchi is shown in Fig. 2. To ease the estimation of the portion size, a spoon and chopsticks are displayed along with the dishes – the serving plate in real size is also shown in the Korean picture book. The weights of portions for 412-1 to 4 are 15, 34, 53 and 72 g, respectively, which increase linearly and show a minimum difference of 25 % between each subsequent portion size.
Fig. 2 An example of photographs for food portion quantification, cabbage kimchi. A colour version of this figure can be found online at http://www.journals.cambridge.org/bjn
Examples of household measures and shapes are shown in the Fig. 3. Dimensions of rice and soup bowls and disposable paper cups are included in the picture book to check the selection with the respondents during the interview. Shape of squared sandwich bread and shapes of pies for bread, pancake, pizza and cake are in real size in the picture book (see Fig. 3).
Fig. 3 Examples of household measures and shapes. A colour version of this figure can be found online at http://www.journals.cambridge.org/bjn
Discussion
The Korean version of GloboDiet is the first Asian version of the international interview-based 24 h dietary recall software that has been already tested for its feasibility, validated and implemented in different epidemiological and surveillance settings in Europe( Reference Crispim, Geelen and Souverein 13 , Reference Crispim, Geelen and Siebelink 33 – Reference Huybrechts, Geelen and de Vries 37 ). To develop the Korean version, the overall structure of databases did not require any change. This confirmed the highly flexible structure of GloboDiet and its major potential as reference methodology for international dietary settings. Indeed, knowing that other Korean neighbouring countries share similar (feature of) dietary patterns and habits, this successful first attempt to adapt GloboDiet outside Europe opens many perspectives to expand the use of this IARC-WHO international tool to other Asian countries, more specifically in the framework of the WHO action plans on diet-related NCD.
In the customisation process, several challenges emerged due to lack of local data and difference in cultural background, compared with Europe. First of all, there was a lack of available databases for local coefficients such as density factors, raw-to-cooked and edible part coefficients that are required to get more accurate estimates of individual dietary intakes (as consumed quantities). For food items without locally available specific coefficients, European coefficients from the German and French versions of GloboDiet were copied, in particular, for coefficients such as density, standard percentages of fat/sauce left in the dish because those coefficients are not so different between Europe and Korea. Second, while most of the GloboDiet versions developed so far used Latin, Greek or Cyrillic alphabets, Korean (Asian) characters were used for the first time in this version. To fit to the Korean language, orders of displayed words and sizes of windows were modified. Due to cultural differences, the way of displaying date and name was also modified. Third, photographs of Korean foods and tableware had to be taken for the picture book as most of the pictures used so far in Europe were not suitable. Although existing pictures are shared across European countries, only four of them (i.e. those available for spread on bread) were retained in the Korean version of the picture book. The new pictures represented frequently consumed Korean foods selected by systematic analysis of 24 h recall data of 15 723 participants from the fifth KNHANES. The number of picture series was determined based on previous experiences from implementations of GloboDiet in other countries.
A validation study is planned to be conducted in the near future aiming to evaluate the reliability of the data collected with the Korean version of GloboDiet against independent and recovery biomarkers such as double-labelled water and urinary N, K and Na. In Europe, GloboDiet has been successfully validated with urinary recovery biomarkers( Reference Crispim, Geelen and de Vries 12 , Reference Crispim, de Vries and Geelen 14 ). Two non-consecutive 24 h recalls using the GloboDiet with a food propensity questionnaire, which captures long-term frequency of non-quantitative intakes of infrequently consumed foods, have shown appropriate validity to rank subject according to their fish and fruit/vegetable intakes( Reference Crispim, Geelen and Souverein 13 ). The association between reporting errors on 24 h recall and BMI was also examined( Reference Freisling, van Bakel and Biessy 38 ). Before the validation study, a pilot study has also been planned considering a convenient sample, in order to check the completeness of the GloboDiet databases (e.g. missing foods and recipes). Currently, a nutrient database – supplementary to the Korean version of GloboDiet – for the calculation of nutrient intakes from the data collected using the Korean version of GloboDiet is under preparation for use in this validation study as well as future planned research and surveillance projects that are still under discussion. Once these planned studies are completed, actual implementation of the methodology in different research and surveillance settings in Korea can be foreseen. For example, the GloboDiet can be implemented in epidemiological studies such as cohort studies as well as nutritional monitoring surveys. To use the GloboDiet for children, foods that are frequently consumed among children will be added to the food list, and the picture book will be updated with household measures for children and pictures of added children's foods.
In Korea, two dietary assessment programs are available: CAN-Pro that was developed by the Korean Nutrition Society( 27 ) and Dietary Evaluation System that was developed by the Human Nutrition Laboratory of Seoul National University( Reference Jung, Lee and Kim 39 ). Both of them were developed to be used for entry of open-ended survey results such as food records or 24 h recall while Dietary Evaluation System was especially designed for online 24 h recall interview. Compared with these programs, nutrient intake of subjects should be calculated with interview data collected with the GloboDiet and nutrient database outside of the software. Since collected interview data can be linked to any kinds of database, which is one of the strengths of the GloboDiet, the data can be used for studies of intakes at food level, meal patterns, risk assessment for food safety, dietary guidelines and so on. In addition, a data entry version of the GloboDiet can be used to enter dietary record data collected from children's parents. To assess children's diet, the international recommendations suggest different approaches than those applied for adults: food diaries or food records( Reference Andersen, Lioret and Brants 40 ). The GloboDiet ‘data entry’ module has already been successfully tested to complement the food records used in children for pan-European food consumption survey among infants, toddlers and children( Reference Ocke, Slimani and Brants 17 , Reference Freisling, Ocke and Casagrande 41 ).
Conclusion
In conclusion, the Korean version of GloboDiet, as the first Asian one, is available for future research including validation study and dietary surveillance in Korea embedded within more IARC-WHO international initiatives. This has been achieved with marginal changes in the standardisation of the databases (e.g. additional specific food classification groups) and with no change in the software structure. This confirms that this IARC-WHO international dietary methodology, successfully used in different pan-European epidemiological and surveillance settings, is flexible and robust enough to be customised for other regions worldwide. It is, therefore, anticipated that the implementation of a common dietary methodology to increase data comparability across Asian and other countries, where the rapid nutritional transition is observed, is possible and could be achieved relatively rapidly according to the current IARC-WHO agendas.
Acknowledgements
The authors thank Hwayoung Noh, Viktoria Knaze, Corinne Casagrande, Sunmi Moon, Jiye Shin and Sunju Choi for their technical assistance.
The present study was conducted and financially supported by the IARC, the owner of the GloboDiet methodology, the National Cancer Centre and the National Research Foundation of Korea (NCC1231100, NRF-2012-040-050). The part of work reported in the present study was undertaken during the tenure of a Postdoctoral Fellowship from the IARC, partially supported by the European Commission FP7 Marie Curie Actions – People – Co-funding of regional, national and international programmes (COFUND). All funders had no role in the design and analysis of the study or in the writing of the article.
The authors' contributions are as follows: N. S. and G. N. led the development of the GloboDiet methodology at the IARC; M. K. P. coordinated the project and wrote the article under the supervision of N. S. and G. N. together with the GloboDiet team at the IARC; J. Y. P., G. N., J. K. and N. S. conceived and designed the study; local databases in Korea were collected by staff at the NCC and the IARC under the guidance of the IARC; J. K. supervised all the activities at the NCC; H. Y. P. contributed to data collection and database customisation; M. K. P. and G. N. mainly conducted the adaptation of databases at the IARC; N. S. had primary responsibility for the final content and supervised the study. Working procedures for the development of the GloboDiet versions were developed at the IARC. All authors critically reviewed the article.
The authors have no financial or personal conflicts of interest to declare.
