Original Research Article
12th IFDC 2017 Special Issue – Evaluation of harmonized EuroFIR documentation for macronutrient values in 26 European food composition databases

https://doi.org/10.1016/j.jfca.2019.03.006Get rights and content

Highlights

  • European food composition data are harmonized according to the EuroFIR guidelines.

  • Documentation on methods and references needs to be further completed.

  • Comparability remains limited for protein, carbohydrates, dietary fiber and energy.

  • Standardized recalculation will further increase comparability and usability.

  • Further harmonization steps are planned for datasets in FoodEXplorer™.

Abstract

Using food composition data from multiple countries requires a high level of harmonization between datasets, at food, component, and value documentation level. To achieve this, nutrient values in European food composition databases were documented using EuroFIR guidelines and thesauri. Our aims were to evaluate the impact of harmonized documentation of macronutrients and usefulness for research and/or policy. Data on 28,914 foods and 292,240 macronutrient values from 26 European food composition datasets were extracted from the FoodEXplorer™ web-based tool for simultaneously searching and comparing food composition data. Documentation on most properties describing the nutrient values was complete, however the percentage coded as unknown varied from 14% to 49% for value type, method type, method indicator and acquisition type. Some inconsistencies were found in coding, and documentation on references was incomplete (about 65% missing information). The harmonized manner of data documentation using the EuroFIR guidelines for food, component and value documentation, allows for data exchange. However, comparability of carbohydrate, dietary fiber, protein and energy values remained difficult due to multiple definitions and formulas used or due to lack of details on analytical and calculation methods. Therefore, data on these components cannot be regarded as fully comparable in Europe without further harmonization; care should be taken when using these data for multiple country comparisons. To achieve further comparability within the FoodEXplorer™ tool, information on methods and references used should be completed for each dataset and energy, carbohydrate and protein values should be recalculated in a uniform way. This remains a major challenge for most countries, but is urgently needed.

Introduction

Although food composition data are usually produced and compiled at national level, they are increasingly used in an international context. Multicenter studies on food consumption and nutrient intake, impacts of food reformulation, nutritional labeling, and other research studies usually need data from more than one national food composition dataset. Data are often exchanged between food composition databases (FCDBs), since it is not possible to analyze each food in each country. To exchange food composition data successfully it is necessary that data are compatible and users must be able to fully understand the data from the metadata that provides additional information.

In Europe, procedures for food and value documentation were developed in the past decades in a number of European projects e.g. EUROFOODS and EFCOSUM (Brussaard et al., 2002, Ireland et al., 2002, Schlotke et al., 2000). Based on these achievements, in the European Prospective Investigation into Cancer and Nutrition (EPIC) study, food composition data from ten European countries were harmonized using a standardized documentation approach (Slimani et al., 2000, Slimani et al., 2007). More recently this work was continued through the EuroFIR projects (EU FP6 & FP7; 2005–2013 (Finglas et al., 2014)).

Using food composition data from multiple countries requires a high level of harmonization, for better comparability of the data, at food, component and value level. Several studies compared national FCDBs, attempted to harmonize datasets and studied the impact on intake calculations (Deharveng et al., 1999, Hakala et al., 2003, Noh et al., 2017, Polacchi et al., 2002, Uusitalo et al., 2011). Protein, carbohydrates, dietary fiber and energy were the macronutrients most frequently identified as being poorly comparable. The main problem was component description in relation to analytical methods and formulas used to analyze and calculate values. To our knowledge, the comparison of nine European FCDBs in 1999 was the most recent inventory including a considerable number of European datasets. From this comparison, it was concluded that more agreement was needed on component definitions, methods of analyses and modes of expression, as well as more documentation with respect to data sources, sampling procedures, analytical methods and formulas used to calculate components and recipes (Deharveng et al., 1999). These results were confirmed by an additional inventory among other European countries in the EFCOSUM project (EFSOCUM-group, 2001). Only a few studies were found that compared intake calculations using European FCDBs directly and mostly adaptations were applied to harmonize datasets prior to calculation and results in general were comparable (Hakala et al., 2003, Julián-Almárcegui et al., 2016, Uusitalo et al., 2011).

In the EuroFIR projects, a generic quality management approach was developed to assure the quality of the compilation process. All steps from data production and data entry, through data aggregation, compilation and control, to data dissemination were documented and agreed upon by compiler organizations (Westenbrink et al., 2009). EuroFIR guidelines were further developed and harmonized (later used as basis for the CEN (European Committee for Standardization) Food Data Standard, 2012 (Becker, 2010)) for food, component and value documentation. The EuroFIR guidelines were reviewed and agreed upon by all project partners and used in their national FCDBs. Thesauri including codes for documentation, based on abovementioned international recommendations were further developed and applied (Becker et al., 2007, Becker et al., 2008, EuroFIR, 2018b, Møller et al., 2008). European compiler organizations agreed on a standardized recipe calculation procedure and food description was harmonized using the LanguaL food description system (LanguaL, 2018). The EuroFIR guidelines enable detailed description of the following main entities: foods, components, values, recipes and references, each with a varying number of properties to document. Mandatory properties for value description are food identifier, component identifier, unit, matrix unit (or mode of expression), value type, acquisition type, value reference, method type and method indicator (Table 1). In addition, details on sampling, analytical methods and quality indicators can be documented. Compiler peer reviews were organized as an achievable alternative to compiler certification (Westenbrink et al., 2016) and demonstrated that European food composition data compilers had made good use of EuroFIR guidelines and thesauri and were working toward improved harmonization of processes wherever possible (e.g. (Biringen Loker et al., 2011, Black et al., 2011, Gurinović et al., 2016, Korošec et al., 2013, Macháčková et al., 2013, Martinez-Victoria et al., 2015, Oliveira et al., 2010, Porubská et al., 2014). Training and support was given to individual compilers by more experienced compilers via a series of workshops and online resources provided in the projects and continued by the EuroFIR Association. Harmonized food composition datasets from 26 European countries were collated in the online FoodEXplorer™ tool (EuroFIR, 2018a), which enables users to simultaneously search and compare food composition data from multiple countries. The FoodEXplorer™ tool is accessible for EuroFIR members (and other users by agreement) from the EuroFIR website, and comprises food composition data from national European and some non-European food composition datasets. The tool allows for simple searches based on food names and advanced searches using the value documentation properties and ranges of nutrient content. A subset of the documented national datasets was coded with FoodEx2 classification codes and delivered to the European Food Safety Authority (EFSA) in 2013 to be used for intake assessment (EFSA, 2015, Roe et al., 2013). Non-European FCDBs currently included in FoodEXplorer™ are from USA, Canada, New Zealand and Japan, each developed and maintained according to their own guidelines and compilation processes.

The aim of this work was to evaluate the documentation of macronutrient values from 26 European datasets collated in the FoodEXplorer™ tool following a sustained period of harmonization activities by European compilers (2011–2018). In particular, attention focused on the extent to which the value documentation of the national datasets was complete, correct and compatible with the EuroFIR guidelines and thesauri. The work also considered whether there had been improvements in the comparability of protein, carbohydrates, dietary fiber and energy values, which were previously reported as not being easily comparable.

Section snippets

Databases

In 2017, the European national food composition datasets (n = 26) for which nutrient values were documented according to the EuroFIR guidelines and thesauri, were extracted from FoodEXplorer™. The datasets, generated between 2005 and 2016 and uploaded into FoodEXplorer™ from 2013, are listed in Table 2. Italy has two national datasets and datasets from Germany and Estonia were not provided. The coverage from Central and Eastern European countries is limited; although work was done on value

Food and component coverage

The total number of foods selected in 26 datasets was 28,914 (ranging from 77 foods in the smallest Czech Republic (CZ) dataset to 3424 foods in the largest UK dataset). The 2009 CZ dataset did not include some key foods, e.g. meat and fish products, but was updated later in 2017 to include about 750 foods.

Fourteen macronutrients were selected, yielding 292,240 individual macronutrient values. Table 4 shows the number of foods and the coverage of macronutrients per national dataset, which on

Conclusions and recommendations

A major achievement, compared to the situation before the EuroFIR project started, is that most national FCDBs in Europe are now documented in a harmonized manner according to the EuroFIR guidelines. Food composition data in Europe are in general comparable using the EuroFIR value documentation system. Still some of the current datasets in FoodEXplorer™ do not give all the details needed due to incomplete documentation. It is recommended to further complete documentation of all mandatory

Disclaimer/conflict of interest

Susanne Westenbrink, Agi Kadvan, Mark Roe, Barbara Koroušić Seljak and Paul Finglas are involved in producing their national food composition tables.

Acknowledgements

The authors gratefully acknowledge the work of the national food composition database organizations in Europe for their work to document their data. The development of the value documentation guidelines and the actual value documentation was completed in part with funds from the EU FP6 Network of Excellence ‘European Food Information Resource (EuroFIR)’ under the EU 6th Framework Quality and Safety Program (FP6-513944), and continued under EuroFIR Nexus under the 7th Framework Program

References (44)

  • H. Noh et al.

    Improving standardization of national nutrient databases for use in international settings: a Korean proof of concept

    J. Food Compos. Anal.

    (2017)
  • W. Polacchi et al.

    Derivation of nutrient values for FAO statistical databases

    J. Food Compos. Anal.

    (2002)
  • J. Porubská et al.

    The Slovak national food composition database: new management system DaRiS

    J. Food Compos. Anal.

    (2014)
  • F. Schlotke et al.

    EUROFOODS recommendations for food composition database management and data interchange

    J. Food Compos. Anal.

    (2000)
  • N. Slimani et al.

    Standardization of food composition databases for the European Prospective Investigation into Cancer and Nutrition (EPIC): general theoretical concept

    J. Food Compos. Anal.

    (2000)
  • U. Uusitalo et al.

    Food composition database harmonization for between-country comparisons of nutrient data in the TEDDY Study

    J. Food Compos. Anal.

    (2011)
  • S. Westenbrink et al.

    Dietary fibre: Challenges in production and use of food composition data

    Food Chem.

    (2013)
  • S. Westenbrink et al.

    Food composition databases: the EuroFIR approach to develop tools to assure the quality of the data compilation process

    Food Chem.

    (2009)
  • S. Westenbrink et al.

    EuroFIR quality approach for managing food composition data; where are we in 2014?

    Food Chem.

    (2016)
  • W. Becker

    Towards a CEN standard on food data

    Eur. J. Clin. Nutr.

    (2010)
  • W. Becker et al.

    Proposal for structure and detail of a EuroFIR standard on food composition data. II. Technical Annex: D1.8.19

    (2008)
  • W. Becker et al.

    Proposal for structure and detail of a EuroFIR standard on food composition data. I. Description of the standard

    (2007)
  • Cited by (11)

    • Comparative analysis of dietary fibre determination by AOAC 991.43 and AOAC 2011.25 for frequently consumed foods in Slovenia

      2022, Food Chemistry
      Citation Excerpt :

      However, there are variations in these dietary fibre intake values between different countries, such that estimations range from 12 g/day to 25 g/day for women, and from 13 g/day to 27 g/day for men (EC, 2020), estimated dietary fibre intake for Slovenian adult population is 20.9 g per day (Seljak et al., 2021). Those estimations will also now be subject to change, as they were based on data from food composition databases where the dietary fibre measures were based on the AOAC 985.29 and AOAC 991.43 methods, and therefore did not include LMWSDF (Westenbrink et al., 2013; Zielinski and Rozema, 2013; Phillips et al., 2019; Westenbrink et al., 2019). As an example, a theoretical meal consisting of 100 g white bread, 200 g soy drink and 200 g apple would contain 7.2 g dietary fibre determined with the AOAC 991.43 method, whereas using the AOAC 2011.25 method, the same meal would contain 19.2 g dietary fibre.

    • Analysis of methodological components and available resources in Costa Rica to generate food composition data

      2022, Journal of Food Composition and Analysis
      Citation Excerpt :

      For the quantification of carbohydrates, an important limitation was found both at the international and national level, since there are many variations due to the use of different definitions and multiple analytical methods and formulas. Westenbrink et al. (2019) commented that to achieve standardization and comparability, more consensus is needed on definitions and analytical and calculation methods (Westenbrink et al., 2019). BeMiller (2017) also agrees with this, stating that definitions of carbohydrate types are not always the same, and analytical methods do not always measure exactly what is included in the definition, leading to controversy over what should be measured and as.

    • Documentation of aggregated/compiled values in food composition databases; EuroFIR default to improve harmonization

      2021, Journal of Food Composition and Analysis
      Citation Excerpt :

      Food composition databases from 36 countries, 26 of which were European, were collated in the EuroFIR FoodEXplorer tool, which allows all datasets to be searched simultaneously (EuroFIR, 2019a). In a recent study (Westenbrink et al., 2019), we evaluated the value documentation of macronutrients in the 26 European datasets and showed that standardized value documentation using predefined thesauri was achieved, which is important for data comparison and data exchange. However, we also concluded that due to differences in documentation observed between datasets, full comparability was not guaranteed.

    • Traditional Balkan foods: Future outlook

      2021, Nutritional and Health Aspects of Food in the Balkans
    View all citing articles on Scopus

    This paper was originally submitted as an oral presentation at the 12th International Food Data Conference held from 11 to 13 October 2017 in Buenos Aires, Argentina.

    View full text