Critical evaluation of folate data in European and international databases: Recommendations for standardization in international nutritional studies

3.1 General

The main folate vitamers and entities, together with the currently used EuroFIR component identifiers (http://www.eurofir.net/eurofir_knowledge/eurofir_thesauri) and INFOODS tagnames (http://www.fao.org/infoods/tagnames_en.stm), are given in Table 1. Recent harmonization between the systems has resulted in the preferred use of “Total Folate” (“FOL”), “Folic acid, synthetic” (“FOLAC”) and “Folate, food” (“FOLFD”; EuroFIR only). The use of the terms “Folate, bound” and “Folate, free” have been discontinued by EuroFIR as these terms have been considered ambiguous and dependent of the method used for deconjugation before determination by microbiological assay (MA).

The overview of selected databases is given in Table 2. Abbreviations are used to refer to these databases in this paper. The countries selected are representative of a wide cross-selection of different European geographical regions. The Slovak database SKA04 mostly contains historical data. New editions of the Dutch, New Zealand, Slovak and US databases were available online after the inventory (late 2009–2010), but were not included in this evaluation. The new Dutch database offers new analytical folate data, revised folate data, and adapted terminologies. The new online Slovak database also has adapted terminologies. The changes in terminologies are discussed in this paper. The New Zealand database contains updated information on fortified foods, differentiates fortified foods from non-fortified foods and industry data from analytical values. The US database has a new food group, updated information on fatty acids, and more vitamin C values. Moreover, the Spanish food database was not included in the inventory because the folate data had not been fully evaluated and published.

The year refers to either the publishing year of the most recent (national) food composition table (DK09, FI08, FR08, DE05, GR04, ITI08, NL06, NZC06, NZF06, NO06, PL05, SE08, CH08, UK08 and US08), the year of the most recent update of the folate database (ITD08), the computerized management database (SKA04) and food label database (SKD09).

All databases are published in the official language and all, except CH08, offer an English translation. There is a trend towards providing databases online as software packages, Excel-files, Access-files, ASCII-files, PDF-files, XML formats, and/or htm(l) (downloadable or not). Most of the databases are already available online. At the time of the inventory, four were only available in print, on CD-ROM and/or disk (NL06, NZC06, PL05 and SKD09). Currently, the new releases of the Dutch and New Zealand databases, as well as the Polish database, are also available online. The majority of the online databases are freely accessible. Some of the databases require a license agreement (DE05 and NZF06), or a fee (PL05).

A summary of database completeness for folate values, terminologies, quantification, unit, and mode of expression, and the sources of values, is given in Table 3. The Greek database was excluded for further analysis because its folate values are taken from the McCance and Widdowson's “The Composition of Foods” 5th edition 17.

Folate completeness is high among the majority of databases, assuming that the zero values are not missing values (80–100%). However, the Slovak databases have lower folate completeness (54% in SKA04 and 12% in SKD09). Some databases have no missing values, and therefore 100% completeness, as previously missing folate values were either assigned a zero value or estimated and/or calculated (DE05, ITI08, NZC06, NZF06, PL05, and SE08). The folate database ITD08 contains only analytical data and does not contain any missing values.

Missing values in DK09, FR08, NL06, NO06, SKA04, SKD09, and US08 were reported to be real missing values due to a lack of data, whereas missing values in FI08 and UK08 were reported to be logical zero and/or trace values. The folate data in the CH08 database have yet to be evaluated and at this stage it is not clear what missing values represent.

Zero values and trace values make up to 13 and 10%, respectively, of folate values in the databases. Zero values were defined in the questionnaire as values assigned to food items that are considered not to contain folate and/or values obtained by analytical methods indicating that no folate is present (i.e. below the method LOD). All databases comprise zero values, except the SKD09 database. Trace values are analyzed or calculated values below LOQ, but the ingredients of the food indicate that a small unquantifiable amount is present. Half of the databases do not identify trace values with a specific descriptor, such as “Tr” or “T” (DK09, DE05, ITD08, NO06, SKA04, SKD09, SE08, and US08). In some of these databases, traces are replaced by zero values (DE05 and NO06). However, this is not an issue when these data are used in practice in user data sets, for e.g. intake calculations, because traces are treated as zero values for this purpose.

3.2 Folate unit and mode of expression

All databases express folate values in μg, except the SKA04 database that expresses its values in mg. The mode of expression per 100 g edible portion is applied in all databases, and is substituted by per serving for supplements in FI08, and by 100 mL food volume for beverages in CH08, liquid products in NL06, and alcoholic beverages in ITI08 and UK08. Alternatively, ITD08 additionally reports per serving in all food groups, and US08 per common household measures.

Calculation adjustments such as recipe calculations and raw-to-cooked calculations are applied in most databases. There is no correction for raw-to-cooked conversions in ITD08 because analyses were also performed on cooked foods. Moreover, values from the McCance and Widdowson tables were subtracted by 27% in the NL06 database due to previous Dutch research that indicated that HPLC analyses yields on average 27% smaller folate values compared with MA values 18. However, these values were converted back to their original value in the new release. In addition, some databases use DFE units (ITD08, NL06, and US08). This takes differences in bioavailability after dietary intake of naturally occurring food folates and folic acid fortificant into account, by using a 1.7 multiplier to convert natural food folates into DFE, and 0.5 for folic acid from fortified foods or supplements. Even though this may be a suitable tool to express recommended folate allowances, the use of DFE increases imprecision in food composition data because the estimation of folate bioavailability is rather crude as it is based on data on status biomarkers from a few human trials only 19. Calculation adjustments are not used in SKD08 and could not be confirmed for CH08.

3.3 Folate quantification and folate terminologies

Commonly, folate data in food composition databases derive from the widely accepted MA with Lactobacillus rhamnosus (ATCC7469) and should be presented as “total folate”. The assay is based on measuring the growth of the organism, which responds similarly to all folate forms with up to three glutamate molecules, but not to folate degradation products 20. The common calibrant in the MA is folic acid.

There is also a European (CEN) standard microbiological procedure (EN14131:2003) available that has been validated by inter-laboratory studies 21-25. A validated method for the determination of total folate in cereals including methods for sample preparation has also been approved 26, 27.

Twelve of the 17 databases (Table 3) use MA as the folate quantification method. Compilers were therefore asked to report the amount of MA-determined folate values in the databases. It was stressed to indicate the origin of these values in order to be able to distinguish analyzed MA folate values from other MA folate values, such as estimates, calculations, values taken from literature, and borrowed values, which are not sufficiently documented. The reported percentages therefore represent minimum percentages due to the exclusion of these other MA folate values. The ITD08 database was the only database that was reported to have 100% (in-house) analyzed MA folate values for all food items.

The databases FI08, FR08, DE05, ITD08, NL06, UK08, and US08 report some data from chromatographic folate quantification. HPLC, LC-MS, or more specific LC-MS/MS methods allow quantification of individual folate forms, but none has yet been officially approved 25, 28-31. HPLC methods are limited by their dependence on commercial folate standards and usually folates are quantified as monoglutamates after enzymatic deconjugation. Few methods are published for the quantification of folate polyglutamates 32, 33. Folic acid is usually quantified using HPLC with UV or electrochemical detection 34, 35 or by LC-MS 36-38. Procedures for the sample preparation and quantification of one of the predominant dietary folate forms (5-methyltetrahydrofolate) have been standardized in a European method intercomparison study 29. HPLC data should be presented either as concentrations of individual folates (in particular 5-methyltetrahydrofolate) or as a “sum of individual folates, expressed as folic acid”. For the purpose of food composition databases it is recommended to specify which individual folates are included in the sum.

For some food items in a few of the databases, FR08, NZC06, NZF06, and SE08, radioprotein-binding methods have been used. Radioprotein-binding methods have been applied with limited success for food analysis. The folate binder in these, for clinical purposes automated, competitive binding assays possesses different affinity to the individual dietary folate forms, and it was recommended to apply this method only for foods containing mainly 5-methyltetrahydrofolate 39.

Most selected databases provide one generic terminology, such as “folate(s)” or “total folate(s)”. Some of those databases provide an additional terminology, corresponding to other folate entities, for a limited amount of food items. The DK09 database provides additional “free folate” values. The ITI08 and NZF06 databases also refer to “folic acid” in a separate note (ITI08) or in an Appendix (NZF06). The UK08 database refers to “5-CH₃-H₄folate”, determined by HPLC, for 1% of food items. Moreover, the CH08, SKA04, and SKD09 databases use “folic acid” as the generic terminology. This terminology should only be used to represent the synthetic form of folate, but it also includes naturally occurring folate in the SKA04 and CH08 databases. However, the new online Slovak database provides “total folate” and “folic acid” terminologies.

Changes in methods and/or terminology may occur over the time period that a database is compiled and consequently the chosen generic terminology does not always correspond to the specific folate entity.

Four databases provide multiple terminologies for a larger amount of food items, either referring to folate bioavailability (“Dietary Folate Equivalents” or “DFE”), the type of analysis used such as MA (“total folate”), and HPLC (“total folates” as the sum of HPLC-derived entities, such as “5-CH₃-H₄folate”, “5-HCO-H₄folate”, “H₄-folate”, and “10-HCO-PteGlu”; and the synthetic “folic acid”), and/or a calculation that subtracts synthetic folic acid from total folate (“food folate”). The FI08 database distinguishes between various folate vitamers, which were all analyzed by HPLC, and refers to the sum of individual folates as “total folates”. However, it is suggested to refer to the sum as “sum of individual folates, expressed as folic acid”. The ITD08 and US08 databases provide mixed terminologies including “total folate”, “folic acid”, “food folate”, and “DFE”. The food folate term in the “DFE” values in both databases was calculated using MA data. The NL06 database includes two terminologies, namely “DFE” and “folic acid”. The latter terminology was used incorrectly in that it referred to naturally occurring folate and synthetic folic acid. However, the new release now provides “folate”, “folic acid”, and “folate equivalents”, respectively, identifying naturally occurring folate, synthetic folic acid and DFE.

3.4 Documentation

One of the key aims of the EuroFIR project was to improve comparability of data within European data sets and allow users access to additional information that could help them decide whether or not values are fit for purpose. Most data sets have traditionally been published in forms that do not easily allow publication of additional information. However, online publication and advances in information management mean that it is increasingly possible to provide the end user with additional information relating to the reliability and limitations of its use.

The key starting point of the EuroFIR data documentation initiative was to provide references for all data to allow further investigation of the data if required. By 2009, compilers from 20 countries had provided references for each published food and wherever possible this was extended to providing references for each individual value.

EuroFIR compilers also agreed to provide full documentation for all newly published data based on the requirements of the EuroFIR technical annex that forms the basis of the proposed CEN standard CEN/TC387 on “Food composition data – Structure and interchange format”. The standard allows documentation of information relating to areas including food description, sample, component, method specification, value, reference, and data quality. The level of detail that can be documented is very comprehensive and documentation properties are split into mandatory properties that are required as they build the core set of documented data that are needed to properly identify and describe the food composition data and optional properties that provide additional information describing the data and thereby assisting the user to assess them.

A system for assessing quality of data from laboratory analytical reports and scientific publications has been developed based on information included in the documentation standard (Oseredczuk, to be published). The EuroFIR e-search facility (http://www.eurofir.net/eurofir_aisbl/membership_benefits/eurofir_esearch_facility) allows data sets to be searched either on the basis of food name or Langual food description identifiers and currently publishes details of the component identifier, unit, mode of expression, value, number of analytical samples, method, and reference.

Details on the availability of folate value documentation are given in Table 4. Compilers reported to have documentation to varying extents for all folate values for DK09, FR08, DE05, ITI08, ITD08, NL06, PL05, SKD09, UK08, and US08. Other databases (FI08, NZC06, NZF06, NO06, and SE08) were reported to have documentation to varying extents for fewer folate values. Two databases were reported to have written documentation in another location (SKA04) or insufficient documentation (CH08) for folate values.

All documented databases provide documentation on the value level. Value documentation of the UK08 database is planned for the near future. Some databases also provide documentation on the food item level for some foods (DK09, FR08, ITI08 and PL05). Folate values in all databases, but DE05, SKD09, NO06, and SE08, were reported to have the main types of documentation (value type, method type, reference type, and acquisition type). Value type is a qualitative description of the value, such as mean, minimum, maximum, and average. Method type is the way a value was obtained, such as analytical result, calculated as recipe or from related food. Reference type describes the publication format, such as “article in book”, “article in journal”, “article in report” or “book”. Acquisition type describes general categories of food composition data sources and methods of data acquisition, such as in-house or affiliated laboratory, independent laboratory, food composition table, peer-reviewed scientific paper and scientific communication (http://www.eurofir.net/eurofir_knowledge/eurofir_thesauri).

All countries, except New Zealand, can furnish bibliographic sources; however, the extent of covered folate values varies between countries. All databases, except SKA04, provide a date or year linked to the folate value, being the date the folate value was entered into the database, date of evaluation, date of analysis, publication date or year associated with an analytical report or scientific article, and/or date when the value was borrowed from another database. Databases that borrow folate data from other databases do not always provide the date when the value was borrowed. However, the publishing year or version of the original database was reported to be found in the bibliographic source for DK09, FR08, ITI08, NO06, and SE08 (not confirmed for PL05 and CH08). Around 50% of the databases offer information on number of samples, minimum and maximum values, and/or value ranges.

Details on method protocols (extraction, deconjugation procedures, specific bacterial strains (MA), analytical quantification method, detection method) and quality assessments (use of certified reference materials, ring tests, and/or accredited laboratories) were not always provided, yet, are essential to evaluate data comparability and reliability. Particularly, MA responds to all folate mono- and diglutamates, while HPLC only quantifies individual folate vitamers for which standards are available and the method was validated for. Moreover, there are no algorithms or other calculations that can be used to convert folate values for another method (e.g. applied in NL06 but not kept in its new release).

Only two databases make use of a quality index or indicator (FI08 and FR08). It is used as a criterion in the aggregation process of folate values for FR08.

3.5 Folic acid-fortified foods in the databases and folic acid fortification status

Results from the inventory show that voluntary fortification is accepted in all countries. The US mandates folic acid to be added to enriched grain products. New Zealand currently authorizes folic acid to be added to bread, except organic breads, but as of 31 May 2012 compliance will be mandatory.

In general, most databases list folic acid-fortified foods. Yet only four databases identify them by means of a symbol or a note (ITD08, NL06, NZC06, and US08); all of which, except NZC06, listing a separate folic acid value. Breakfast cereals are most commonly fortified across countries. Other fortified foods are cereal products, beverages, some dairy products, dietetic foods, clinical and infant formulas, fortified biscuits and bars.

A few national compilers could not confirm the presence of folic acid-fortified foods in their database. The German compiler reported to have no folic acid-fortified foods in their current database. The next update will provide folic acid-fortified foods, only specifically identified as being folic acid fortified when less than 2 vitamins have been added, and without a separate folic acid value.