Food Chemistry 113 (2009) 736–747

Contents lists available at ScienceDirect

Food Chemistry

journal homepage: www.elsevier .com/locate / foodchem

Review

Inventory of phytoestrogen databases

Heidi Schwartz a,*, Gerhard Sontag a, Jenny Plumb b

a Department of Analytical and Food Chemistry, Faculty of Chemistry, University of Vienna, Waehringer Strasse 38, A-1090 Vienna, Austriab Institute of Food Research, Norwich Research Park, Colney, Norwich NR4 6UJ, UK

a r t i c l e i n f o

Article history:Received 23 November 2007Received in revised form 8 September 2008Accepted 13 September 2008

Keywords:PhytoestrogensIsoflavonesLignansDatabasesFoodAnalysis

0308-8146/$ - see front matter � 2008 Elsevier Ltd. Adoi:10.1016/j.foodchem.2008.09.051

* Corresponding author. Tel.: +43 1 4277 52316; faE-mail address: heidi.schwartz@univie.ac.at (H. Sc

a b s t r a c t

In this review, 17 phytoestrogen databases (PE DBs) including three literature compendia, 11 DBs for PEintake assessment in different countries or population groups and three comprehensive DBs for nutritionresearch were compared with respect to several issues, specifically the number of foods and compoundscovered, the data sources, the mode of data expression, the additional information presented and thequality control of the data. The problems encountered in the construction and use of PE DBs (natural var-iability of PE contents, incomplete coverage of foods and compounds) were discussed alongside therequirements of DBs intended for intake assessment or nutrition research. In addition, recommendationswere given on which DBs are best suited for which purpose. The reviewed DBs differ in the date of con-struction, aim, structure and also in comprehensiveness. The greatest number of foods is covered in DBsfor intake assessment based on national food consumption data, whereas most information is given incomprehensive DBs for nutrition research. Presentation of quality assessed data is of increasing impor-tance as new developments in PE analysis and steady production of new analytical data make replace-ment of low quality data possible.

� 2008 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7372. Chronology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7373. Comparison of phytoestrogen databases: recommendations and current status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 737

3.1. Types and uses of phytoestrogen databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7373.2. Selection of foods and compounds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7413.3. Data sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 741

3.3.1. Analytical measurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7413.3.2. Peer-reviewed literature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7423.3.3. Recipe calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7423.3.4. Use of other databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7423.3.5. Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .742

3.4. Mode of data expression. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7423.5. Amount of additional information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7433.6. Quality control of the data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743

4. Selection of the best-suited databases for different purposes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 744

4.1. Suitability of the available DBs for intake assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 744

4.1.1. Isoflavones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7444.1.2. Lignans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .744

4.2. Suitability of the available DBs for nutrition research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 745

5. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7456. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .746References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746

ll rights reserved.

x: +43 1 4277 9523.hwartz).

H. Schwartz et al. / Food Chemistry 113 (2009) 736–747 737

1. Introduction

Phytoestrogens (PEs) are bioactive compounds capable ofinducing a diverse range of health effects. Among these are effectson cardiovascular diseases, hormone dependent cancers, meno-pausal symptoms, bone density, as well as antioxidant, anti-inflammatory and vasodilatory effects (Adlercreutz, 2007; Cassidyet al., 2006; Messina, Kucuk, & Lampe, 2006; Mitjans & Vinardell,2005; Mortensen et al., 2009; Sacks et al., 2006; Westcott & Muir,2003). Depending on their structure, PEs can be divided into flavo-noid and non-flavonoid polyphenols. The main representatives offlavonoid PEs are isoflavones and coumestans. Lignans and coume-stans are the main non-flavonoid PEs.

Isoflavones, which occur mainly in soybeans, are the most stud-ied class of PEs. The main soy isoflavones are daidzein (daid), gen-istein (gen) and glycitein (gly). In some legumes, formononetin(formo) and biochanin A (bio A), the methoxylated forms of daidand gen, can be found alongside with coumestrol (coum), the maincoumestan. In unprocessed soybeans, isoflavones occur mainly asmalonyl-b-glucosides and b-glucosides. Upon processing involvingelevated temperatures, malonyl forms are degraded to acetylforms, b-glucosides and, depending on the extent and kind of heattreatment, to aglucones (Coward, Smith, Kirk, & Barnes, 1998). Thebest dietary isoflavone sources are traditional soy foods, new gen-eration soy foods (for example, soy burgers or soy desserts), anddaily consumed basic foods including bakery goods and meat prod-ucts, to which soy flour or soy protein had been added.

Lignans are widely encountered in the plant kingdom. Highestcontents (up to the mg/g range) have been found in flaxseed andsesame seeds. Lower concentrations (medium ng/g–lg/g range)have been determined in vegetables, legumes, cereals, nuts, fruitsand beverages as, for example, tea, coffee and wine. According tothe current standard of knowledge, the main dietary lignans aresecoisolariciresinol (seco), lariciresinol (larici), pinoresinol (pino),medioresinol (medio), syringaresinol (syringa), 7-hydroxyma-tairesinol (7-OH-matai), matairesinol (matai) and the sesamelignans sesaminol, sesamolinol, sesamin and sesamolin (Achouri,Boye, & Belanger, 2005; Milder, Arts, van de Putte, Venema, &Hollman, 2005; Moazzami & Kamal-Eldin, 2006; Penalvo, Adlerc-reutz, Uehara, Ristimaki, & Watanabe, 2008; Penalvo, Heinonen,Aura, & Adlercreutz, 2005; Smeds et al., 2007; Thompson, Bou-cher, Liu, Cotterchio, & Kreiger, 2006). These lignans occur in dif-ferent native forms, sometimes of unknown complexity. Forinstance, seco, the best studied lignan, occurs as bio-oligomerconsisting of seco diglucoside (SDG) units linked by 3-hydroxy-3-methylglutaric acid in flaxseed (Kamal-Eldin et al., 2001). Re-cent analytical findings indicate that other lignans occur partlyesterified, too (Milder et al., 2005; Schwartz & Sontag, 2006;Smeds et al., 2007). This may depend on the plant, the plant vari-ety, and on environmental factors.

The isoflavone and lignan patterns and contents in foods areinfluenced by a variety of factors, which include natural variability,variability caused by processing and variability introduced by theuse of different ingredients and formulations (Eliasson, Kamal-Eldin, Andersson, & Aman, 2003; Lee et al., 2003; Moazzami &Kamal-Eldin, 2006; Mortensen et al., 2009; Setchell et al., 2001).This high variability of PE contents in foodstuffs of the same typeis one of the main problems in creating phytoestrogen databases(PE DBs), independent of their structure, data sources and uses.

In the following, a short chronology of the development of PEDBs will be given, introducing the individual DBs. Then, the indi-vidual PE DBs will be compared with respect to the selection offoods and compounds, the data sources, the mode of data expres-sion, the amount of additional information and the quality controlof the data. At the same time, difficulties in the compilation of PEDBs will be highlighted, and advantages and limitations of the dif-

ferent approaches will be discussed. Finally, recommendations willbe given on which DB should be chosen for which purpose.

2. Chronology

In order to relate the consumption of PEs to health effects, it isessential to have a thorough overview of the PE contents in thefoods consumed by the population under study. Early data collec-tions summarising PE contents in selected foodstuffs from differ-ent references were aimed mainly at identifying dietary PEsources, summarising existing values, comparing analytical meth-ods used and giving recommendations for future analyses (Mazur& Adlercreutz, 1998; Meagher & Beecher, 2000; Reinli & Block,1996), or at assessing the dietary intake of a population sub-group or of participants in epidemiological studies by means ofa food frequency questionnaire (FFQ) containing only a limitednumber of foods (de Kleijn et al., 2001; Fink, Steck, Wolff, Kabat,& Gammon, 2006; Horn-Ross et al., 2000; Pillow et al., 1999). Thefirst isoflavone DB, which was freely accessible to the researchcommunity and to the public and could be used independentlyof the paper describing its construction, was released in 1999(US Department of Agriculture, 2002). However, only a limitednumber of food items were covered at that time. In 2001, theBASIS DB, a DB for bioactive compounds (including isoflavones,lignans and coumestrol) in �300 food plants, became availableon CD-ROM. Two years later, reports of two further PE DBs werepublished, one of them containing PE values for �800 (Kiely,Faughnan, Wahala, Brants, & Mulligan, 2003), the other for�1400 (Valsta et al., 2003) foods. The main aim of these and ofthe following DBs was intake assessment in different countriesand populations. In 2005, a free access internet-based compre-hensive isoflavone DB comprising �6000 foods was produced(Ritchie, n.d.; Ritchie et al., 2006), and the first DB containing val-ues for the lignans larici, pino, seco and matai was published(Milder et al., 2005). One year later, a DB providing values forthe major isoflavones, lignans and coumestans in foods, accordingto the present standard of knowledge, was established by thesimultaneous analysis of the three classes of PEs in the samefoods (Thompson et al., 2006). Most recently, the inclusion of lig-nan (seco and matai) values into an existing food composition ta-ble listing �1500 foods was reported (Blitz, Murphy, & Au, 2007).In addition, one further comprehensive DB with multiple usescontaining detailed information about the food and data genera-tion is currently being developed (Scalbert, 2007) and one isbeing updated (US Department of Agriculture, 2002). Finally, theBASIS DB is currently being both enlarged to the EuroFIR BASISDB and updated (Gry et al., 2007). Table 1 gives an overview ofthe reviewed PE DBs, sorted by their main aims and uses.

3. Comparison of phytoestrogen databases: recommendationsand current status

3.1. Types and uses of phytoestrogen databases

PE DBs can be divided into three groups according to their mainaims: literature compendia aiming at identifying PE sources, pro-viding a collection of available data and helping to prioritise futureanalyses; DBs for dietary intake assessment either by means offood frequency questionnaires (for population sub-groups or par-ticipants of epidemiological studies) or by means of national foodconsumption studies and DBs providing sufficiently detailed addi-tional information on the food sample, the analytical method usedto generate the data and on the quality of the individual values sothat they can be used as resource for the regulatory affairs sector,the food industry and scientists.

Table 1Overview of phytoestrogen databases

Reference Type of data collection;main aim; accessibility

Number and types of foods Data sources; aggregation ofdata

PEs covered; expression of thedata

Additional information; qualitycontrol

Time period covered

Reinli and Block (1996) Literature compendium;identification of PEcontaining foods; all dataprovided in the article

�90 different foods or foodsdifferently processed comprising160 indiv. food items (trad. soyfoods, new gen. soy foods, soyingredients, legumes); separatevalues given for differentvarieties (e.g. soybean)

Peer-reviewed literature;aggregated data for 13 popularAsian foods in the American diet(mean of data for P2 lots orbrands given in 1 article, medianof data for P2 types or brands ofa similar food from differentpapers) and non-aggregated data

Daid, gen, formo, bio A; lg/gaglycone equivalents, as-is basis

Processing, cultivar or brand,data source, sample preparation,method of analysis (extraction,hydrolysis), use of IS, CV,moisture content, number ofanalyses (not stated whetheranalysis of aliquots of the samehomogenised sample unit or ofdifferent sample units) for non-aggregated data from one paper,number of foods analysed foraggregated data; QC: conflictingresults excluded

Latest ref. from 1995

Mazur (1998) Literature compendium;report on quantitativeresults for PEs in plantfoods; all data provided inthe article

65 (legumes, cereals, oilseedsand nuts, fruits, vegetables, non-alcoholic beverages, wine)

Laboratory analysis by theauthors, in part published inpeer-reviewed journals; non-aggregated data; range fordifferent cultivars of legumes

Daid, gen, formo, bio A, coum,seco, matai; lg/100 g aglyconeequivalents, DW basis

Latin name, grape variety,country and type of wine in thecase of wines, ref. to the datasource, analytical method;analytical QC

Latest ref. from 1998

Meagher and Beecher(2000)

Literature compendium;compilation of the lignancontent of food groups,comparison of results fromchemical analysis with datafrom in vitro fermentation;all data provided

55 different foods (cereals,legumes, oil seeds, fruits,vegetables, beverages and someproc. foods), for some foodsmore than 1 value reported

Peer-reviewed literature; non-aggregated data

Seco, matai, seco plus matai, end,enl, end plus enl; lg/100 gaglycone equivalents, DW basis

Ref. to the data source, analyticalmethodology; no QC

Latest ref. from 1998

Pillow et al. (1999) DB for use with a FFQ(Health Habits and HistoryQuestionnaire); all dataprovided

Isoflavone values for 28 foods(trad. soy foods, new gen. soyfoods, legumes), lignan valuesfor 25, mammalian lignan valuesfor 43 in part proc. plant foods

Literature, calculated usingstandard recipes from USDA ormanufacturers’ info; aggregateddata (mean if a single ref.reported multiple analysis of thesame brand or specific type offood, median if values for anindiv. food item were fromdifferent refs. or for differentbrands or types in 1 ref.)

Daid, gen, formo, bio A, coum,seco, matai, end, enl; lg/100 gaglycone equivalents, weightbasis not stated

Ref. to the data source; no QC Latest ref. from 1998

Horn-Ross et al. (2000) DB for use with a FFQdeveloped to assess therelationship between PEexposure and cancer risk;all data provided

42 foods containing PEs (trad.soy-based foods, proc. foodswith added soy flour or soyprotein, vegetables, legumes,fruits), 78 foods with zero values

Laboratory analysis aftersampling according to anelaborate sampling plan; non-aggregated data

Daid, gen, formo, bio A, coum,seco, matai; lg/100 g aglyconeequivalents, as-is basis (edibleportion)

No additional information; noQC

Paper submitted inMarch 1999

de Kleijn et al. (2001) DB for use with a FFQ(Willet FFQ); all dataprovided

Isoflavone values for 27 foods,lignan values for 63 foods (proc.and non-proc. plant foods), �65foods with zero values

Literature, previouslyunpublished data, estimationusing the PE content of a similarfood; non-aggregated data

Daid, gen, formo, bio A, coum,seco, matai; mg/100 g aglyconeequivalents. as-is basis

Moisture content; no QC Latest ref. from 1999

Fink et al. (2006) Expanded flavonoid DB foruse with a modified BlockFFQ for flavonoid and lignanintake assessment in theLong Island breast cancerstudy; all data provided

50 FFQ items representing 100proc. and non-proc. foods;isoflavone values for 19 foods,lignan values for 39 foods

USDA DB, literature, other DBs,published laboratory data;aggregated data (median)

Total isoflavones, total lignans;mg/100 g aglycone equivalents,weight basis not stated

No additional information;quality assessment according tothe USDA expert qualityassessment scheme

Latest ref. from 2002

(continued on next page)

738H

.Schwartz

etal./Food

Chemistry

113(2009)

736–747

Table 1 (continued)

Reference Type of data collection;main aim; accessibility

Number and types of foods Data sources; aggregation ofdata

PEs covered; expression of thedata

Additional information; qualitycontrol

Time period covered

Kiely et al. (2003) MS Access� based electronicDB; PE intake assessment indifferent countries; nolonger available to thepublic

Isoflavone values for 791 foods,lignan values for 158 foods (300foods commonly consumed inEurope)

Peer-reviewed literature,analytical measurement; non-aggregated data

Daid, gen, formo, bio A, totalisoflavones, coum, seco, matai;lg/100 g aglycone equivalents,as-is basis (edible portion)

Ref. to the data source, ref. ID,analytical method, hydrolysis Y/N, internal standard Y/N, QC Y/N,number of samples; QC prior toselection of papers

Latest ref. from 2000

Valsta et al. (2003) Electronic DB; PE intakeassessment; for internaluse, not available to thepublic

Isoflavone and lignan values for�1400 foods listed in the FinnishNational food composition DB

Laboratory analysis, peer-reviewed literature, estimationusing isoflavone contents ofcomparable foods, recipecalculations, estimation basedon information from the foodindustry; non-aggregated data

Daid, gen, seco, matai; lg/100 gor lg/dl aglycone equivalents,as-is basis

Ref. to the data source, qualitycode; QC of the data in the DBaccording to four criteria

Latest ref. from 2001;data will be updatedin case of futureresearch needs

Ritchie et al. (2006) Electronic DB; isoflavoneintake assessment;accessible on the internet

�6000 proc. and non-proc.foods, of these �960 with non-zero values

Peer-reviewed literature,estimation using isoflavonecontents of similar foods, recipecalculations, calculations basedon nutrient info fromsupermarket chains or from UKmanufacturers; non-aggregateddata

Total daid plus gen; lg/100 gaglycone equivalents, as-is basis

McCance & Widdowson code,analytical method, ref. to theoriginal data source, method ofobtaining the data (estimated,calculated, recipe calculation,measured); QC at 2 levels: (1)prior to inclusion of values intothe DB and (2) test of theestablished DB

Literature datamainly from 2000;validation of the DBpublished in 2004

Milder et al. (2005) DB for evaluation of thehealth effects of lignanintake; all data provided

109 (nuts and seeds, grainproducts, vegetables, legumes,fruits, few proc. foods, alcoholicand non-alcoholic beverages)

Laboratory analysis of compositesamples (edible parts of plantfoods); non-aggregated data

Seco, matai, larici, pino, totallignans; lg/100 g aglyconeequivalents, as-is basis (edibleportion)

Latin name, processing method,moisture content; analytical QC

Paper submitted in2004

Thompson et al. (2006) DB for PE intake estimationin epidemiological andclinical studies; all dataprovided

121 (soy products, legumes, nutsand seeds, vegetables, fruits,grain products, meat products,processed multi ingredientfoods, alcoholic and non-alcoholic beverages)

Laboratory analysis of compositesamples (edible parts of plantfoods); non-aggregated data

Daid, gen, gly, formo, totalisoflavones, coum, seco, matai,larici, pino, total lignans, totalPEs; lg/100 g and lg/serving inaglycone equivalents, as-is basis

Brand names, country of origin,special ingredients, color,processing method, mean andstandard deviation of the foodsin one food group; no report ofQC

Paper submitted inDecember 2005

Blitz et al. (2007) Proprietary DB; lignanintake assessment; forinternal use, not accessibleto the public

1332 proc. and non-proc. foods Peer-reviewed literature,imputation based on similarfoods, recipe calculation,calculation based on the fibrecontent; aggregated data (mean)

Seco, matai; lg/100 g aglyconeequivalents, as-is basis

Processing method, ingredients,method of assigning the lignanvalue, ref. to the data source;quality of values judged byreported recovery and precisionof the analyses

Data were compiledin early 2003; datawill be updated

Park et al. (2007) Electronic DB; isoflavoneintake estimation inKoreans; all data given inthe article

142 Korean soy foods Literature (25 food items), recipecalculation and adaptation fromthe values of similar food items(98 foods), USDA DB (19 foods);aggregated analytical data(median)

Daid, gly, gen, total isoflavones;mg/100 g aglycone equivalents,as-is basis

Food code in the Korean NutrientDB, processing, type of datasource, but not ref. to the datasource, moisture content,confidence code; QC: (1)exclusion of articles withinsufficient information orinappropriate sampling and (2)QC according to the USDA expertsystem. Aggregation of data fromreliable methods only

Articles betweenJanuary 1990 andMarch 2004

H.Schw

artzet

al./FoodChem

istry113

(2009)736–

747739

US Department ofAgriculture (2002)

MS Access� based electronicDB; nutrition research;freely accessible on theinternet

128 (legumes, infant formulas,grain products, trad. and newgen. soy products)

Laboratory analysis, peer-reviewed literature; aggregateddata (mean)

Daid, gen, gly, total isoflavones;mg/100 g aglycone equivalents,as-is basis (edible portion)

Nutrient data bank number,mean, standard error of themean, total number of means orindiv. values, min and max value,confidence code, data source;only data meeting minimumrequirements were included intothe DB; expert quality evaluationsystem

Latest ref. from 1998.However, the DB iscurrently beingupdated

Scalbert (2007) Electronic DB; creation of acomprehensive foodcomposition table forpolyphenols; DB will belinked with European foodcomposition DBs and will befully accessible to the public

Ongoing project Peer-reviewed literature; non-aggregated data in the DB,aggregated data in foodcomposition tables

All PEs from peer-reviewedliterature; aglycones and nativeforms as given in the literature;as-is basis

Variety, origin of the sample,data source, sample preparationmethod, analytical method,standards used, number ofsamples. Food compositiontables: compound family, mean,standard deviation, min value,max value, number of dataaggregated; quality assessmentbefore inclusion into the DB, noquality scores

Ongoing project

Gry et al. (2007),EuroFIR BASIS

Internet-deployed DB;resource for regulatoryaffairs and risk authorities,epidemiologists,researchers and productdevelopers within the foodindustry; accessibility indiscussion

�300 plant foods andunspecified number of proc.plant foods, ongoing project

Peer-reviewed literature; non-aggregated data

All PEs from peer-reviewedliterature; aglycones and nativeforms as published; units andweight basis as given in theoriginal article

Detailed info on the food plant,processing, sampling, analyticalmethod, analytical standardsource, variability (min, max,standard deviation, meanstandard error) and quality ofthe data; expert qualityassessment scheme

Ongoing project

PE: phytoestrogen, daid: daidzein, gen: genistein, gly: glycitein, formo: formononetin, bio A: biochanin A, coum: coumestrol, seco: secoisolariciresinol, larici: lariciresinol, pino: pinoresinol, matai: matairesinol, end: enterodiol, enl:enterolactone, DB: database, FFQ: food frequency questionnaire, QC: quality control, DW: dry weight, CV: coefficient of variation, IS: internal standard, Y: yes, N: no, min: minimum, max: maximum, ref.: reference, indiv.:individual, info: information, trad.: traditional, gen.: generation, and proc.: processed.

740H

.Schwartz

etal./Food

Chemistry

113(2009)

736–747

H. Schwartz et al. / Food Chemistry 113 (2009) 736–747 741

3.2. Selection of foods and compounds

The number of foods covered in a DB depends on the aim. DBsfor intake assessment in different countries or populations shouldcover as many food items and compounds as possible and, likecomprehensive DBs for nutrition research, should especially coverall basic foodstuffs (used either for direct consumption or as ingre-dients in composite foods) in order to make recipe calculationspossible, which greatly enlarge the number of foods covered andwhich enable expert users to perform their own calculations. In-deed, whereas literature compendia and DBs for intake assessmentof a certain population sub-group (Table 1) generally contain <100in part basic, in part processed food items, some of the DBs for in-take assessment in different countries cover more than 1000 foodsand comprehensive DBs with multiple uses contain a high percent-age of basic foods.

The authors of literature compendia selected the foods forinclusion on the basis of the literature data available at the timeof compilation (Meagher & Beecher, 2000; Reinli & Block, 1996).Mazur (1998) who summarised data from own analyses selectedthe foods for analysis in such a way as to cover a wide range of ba-sic plant foods and some beverages commonly consumed inFinland.

Authors of DBs or food composition tables (FCTs) intended foruse with a FFQ included all foodstuffs listed on the FFQ (de Kleijnet al., 2001; Fink et al., 2006; Pillow et al., 1999). In addition, Pillowand co-workers included further commonly eaten foods and foodswhich were considered top sources of PEs and available at localgrocery or health food stores. Horn-Ross and co-workers (2000) se-lected foods to be put onto their FFQ (created to investigate therelationship between PE exposure and cancer risk) according tothe results of semi-structured interviews with 118 African–Amer-ican, Latina and white women from the California San FranciscoBay Area. In addition, a variety of foods suspected to be goodsources of isoflavones were included and results from the firstanalyses influenced the selection of further foods for analysis.

Foods to be covered in DBs for intake assessment in differentcountries were selected according to national food consumptionsurveys (Milder et al., 2005; Thompson et al., 2006) and accordingto FFQs, dietary recalls and food consumption diaries (Ritchie et al.,2006). In addition to using foods listed on a modified Block FFQ,Thompson and co-workers included foods identified as good PEsources and considered relevant to Western populations. Specificfood types and brands were those most frequently consumed byparticipants in a breast cancer case-control study. Milder et al.(2005) included only foods with an average consumption of >3 gper person and day except for fruits (>1 g) and beverages (>10 g),and foods reported to have high lignan contents even if they didnot meet the consumption criteria. With few exceptions, onlyone-component plant foods were analysed because the lignan con-tent of mixed dishes can be estimated using standard recipes. Blitzet al. (2007) and Valsta et al. (2003) ascribed PE values to foodscontained in the existing FCTs or DBs. Kiely and co-workers(2003) selected foods rich in PEs as well as commonly consumedfood items containing low levels of isoflavones.

Two of the comprehensive DBs for multiple uses contain mainlybasic plant foods, for which literature data are available (Gry et al.,2007; Scalbert, 2007). However, the EuroFIR BASIS DB is planningto include processed foods, for example, soy milk, tofu, miso, soysauce, soy flour and other soy-based foods as well as wines, choc-olate and grain products, for example, bread, cookies, muffins andcereals in the second phase of the EuroFIR project (www.euro-fir.org). The USDA DB contains a variety of basic and processedfoods commonly consumed in the USA.

Of the 17 reviewed PE DBs, nine contain data for both isoflav-ones and lignans, five contain only isoflavone data and three spec-

ialised on lignans. The main isoflavones daid and gen are coveredin all DBs presenting isoflavone data, whereas gly, the third soy iso-flavone, is listed in only four DBs. Values for formononetin andbiochanin are included in eight DBs. The selection of lignans is gov-erned by the amount of data available for the individual com-pounds. Prior to 2005 analytical data for lignans in variousfoodstuffs had been available only for seco and matai so that ofthe 12 DBs containing lignan values only the four newest ones in-clude also larici and pino (Gry et al., 2007; Milder et al., 2005;Scalbert, 2007; Thompson et al., 2006) and medio, syringa as wellas 7-OH-matai (Gry et al., 2007), which have been shown to beequally abundant in foods (Penalvo & Nurmi, 2006; Smeds et al.,2007). The article by Blitz and co-workers (2007) had been submit-ted only a short time after the paper by Milder et al. (2005) had beenpublished, so that values for larici and pino had not been available atthe time of data compilation. Hence, the lignan intake determinedusing earlier created DBs was considerably underestimated.

3.3. Data sources

The more foods to be covered, the more data from differentsources will be required. In the reviewed DBs, data were generatedby laboratory analysis, extracted from peer-reviewed literature,obtained by recipe calculation using ingredient lists and informa-tion from the supermarket chains or from the food industry, ‘bor-rowed’ from other DBs and/or estimated based on the PE contentof similar foods and/or on the PE to fibre ratio. In the following,the advantages and limitations of the individual data sources willbe discussed with examples from the reviewed DBs.

3.3.1. Analytical measurementThe advantages of data generation by laboratory analysis are

that sampling can be performed exactly as required so that repre-sentative data can be obtained and that all PEs of interest can beincluded, ensuring that differences in the PE contents of the inves-tigated foodstuffs are due to natural differences and not due toanalysis by different methods.

The problem both in isoflavone and in lignan analysis is the lackof one commonly used and universally applicable sample prepara-tion method. In isoflavone analysis, methods based on the extrac-tion and analysis of all native forms, e.g. Collison (2008), are usedmore often than hydrolytic methods despite problems like limitedstability of malonyl and acetyl glucosides in extracts and in stan-dard solutions (Griffith & Collison, 2001) and greater danger of coe-lution upon HPLC analysis of all native forms. If only isoflavoneaglucone contents are required, enzymatic hydrolysis with non-specific enzyme preparations like Helix pomatia juice or cellulase,which are capable of cleaving glucosides, and esterified glucosidesis the method of choice (Liggins, Bluck, Coward, & Bingham, 1998;Schwartz, 2008; Thompson et al., 2006). The limitations of acidhydrolysis, the most frequently used hydrolytic technique, are thatthe cleavage of isoflavone glucosides may be incomplete if mildconditions (61 M hydrochloric acid, refluxing for 62 h) are chosen(Penalvo, Nurmi, & Adlercreutz, 2004; Schwartz, 2008) and thathigher molarity of hydrochloric acid or longer incubation timeslead to the partial degradation of gen, e.g. Muellner and Sontag(1999). Base hydrolysis, which is suitable for cleaving ester bondsand therefore for determining both total glucosides and free aglu-cones, is rarely used although an AOAC Official Method has beendeveloped (Official Methods of Analysis, 2000).

Lignan analysis usually includes a hydrolysis step. In early pa-pers, only seco and matai were determined by methods as differentas acid hydrolysis (Liggins, Grimwood, & Bingham, 2000), succes-sive enzymatic and acid hydrolysis (Mazur et al., 1996), alkalineextraction (Eliasson et al., 2003), alkaline extraction plus enzy-matic hydrolysis (Kraushofer & Sontag, 2002) and extraction fol-

742 H. Schwartz et al. / Food Chemistry 113 (2009) 736–747

lowed by base hydrolysis (Johnson, Kamal-Eldin, Lundgren, &Aman, 2000) or enzymatic hydrolysis with Helix pomatia juice(Obermeyer et al., 1995; Setchell, Childress, Zimmer-Nechemias,& Cai, 1999) or b-glucosidase (Horn-Ross et al., 2000).

The limitation of acid hydrolysis is that several lignans are acidlabile, resulting in a number of decomposition products. Enzymatichydrolysis, on the other hand, is not suitable for cleaving complex,in part bio-oligomeric compounds, so that contents obtained byjust enzymatic hydrolysis are likely to be underestimated. Themost frequently employed sample preparation method from2004 onwards for the analysis of most lignans (except the base la-bile 7-OH-matai) includes alkaline hydrolysis or methanolysis forthe cleavage of ester bonds followed by enzymatic hydrolysis forthe cleavage of glycosidic bonds (Milder et al., 2005; Penalvo,Haajanen, Botting, & Adlercreutz, 2005; Thompson et al., 2006).However, most recently a trend-setting article testing differentsample preparation methods for the analysis of a broad range oflignans in cereals and oilseeds has been published (Smeds et al.,2007), stating that alkaline extraction followed by mild acid hydro-lysis and enzymatic hydrolysis gave higher contents than alkalineextraction and enzymatic hydrolysis alone, indicating that sub-stantial analytical method development is still required.

Laboratory analysis was the sole data source of four of the re-viewed PE DBs. Three of these DBs contain data for both isoflav-ones and lignans in the same foodstuffs (Horn-Ross et al., 2000;Mazur, 1998; Thompson et al., 2006) and all three used differentsample preparation methods: Mazur (1998) worked up by enzy-matic hydrolysis followed by the extraction of aglycones, acidhydrolysis of remaining glycosides and combination of both solu-tions prior to further purification (Mazur et al., 1996). Horn-Rossand co-workers employed enzymatic hydrolysis after extractionand Thompson et al. (2006) divided the extracts, performingenzymatic hydrolysis for the determination of isoflavones andalkaline hydrolysis followed by enzymatic hydrolysis for lignananalysis. Determining only lignans, Milder and co-workers optedfor alkaline extraction followed by enzymatic hydrolysis. The het-erogeneity of these methods reflects the general inconsistency inthe selection of sample preparation methods for PE analysis andposes one of the great problems when literature data are includedin PE DBs.

3.3.2. Peer-reviewed literatureAs direct laboratory analysis can be performed only for a lim-

ited number of samples, comprehensive DBs are reliant on peer-reviewed literature as a data source. Indeed, all but the databasescontaining solely own analysis data contain data from peer-re-viewed literature. However, a high proportion of analytical dataoriginates from method development papers, where single sam-ples (sometimes not even properly described) were purchasedin low quantity at a convenient location without sampling planand used to test the applicability of the developed method. Valuesfor these so-called convenience samples are not representative assuch, but can be valuable contributions in DBs, where aggregateddata are presented (see Section 3.4). On the other hand, data in-tended to be included into DBs were sometimes obtained usingoutdated or not validated analytical methods. Hence, literaturedata are to be used with caution and all relevant informationon the samples, sampling, sample handling, the analytical methodand the analytical quality control, which is given in the originalpaper, should be carried over into the DB to facilitate the useand, if necessary, the replacement of the data in the future.

3.3.3. Recipe calculationCalculating PE contents of processed multi-ingredient foods

using recipes, ingredient lists or information from the manufactur-ers has one main advantage over direct laboratory analysis: every

time the recipe or formulation is modified or a different ingredientis used, analytical data of the composite dish become outdated.Calculations, however, can easily be adjusted. Unfortunately, theisoflavone content of soy flour or textured vegetable protein, whichmay be used in the manufacture of processed foods, is highly var-iable (Setchell & Cole, 2003). In addition, research articles often useterms such as soy, soy protein, or even soy supplements withoutspecifying what is meant (Erdman, Badger, Lampe, Setchell, Mes-sina, 2004). Likewise, lignan contents in plant foods used for pre-paring multi-ingredient foods are subject to natural variations.Hence, all kinds of calculations can only be estimations. Despitethat, five of the reviewed DBs rely on recipe calculations as a datasource (Table 1), which explains the great number of foodstuffscovered in three of them (Blitz et al., 2007; Ritchie, n.d.; Valstaet al., 2003).

3.3.4. Use of other databasesWhen data are ‘borrowed’ from other DBs, it should be ascer-

tained that these data are applicable. For instance, soy flour andtextured vegetable protein are used in the production of frequentlyconsumed foods such as white bread, muffins or canned tuna insome countries. Although levels are low (generally <2 mg/100 g to-tal isoflavones), these basic foods are the main source of PEs forsome population groups. Hence, wrongly ascribed values greatlyaffect the outcome of epidemiological studies. In only one of thePE DBs covered in this work data were taken from another DB (Finket al., 2006), and this measure was justified because both DBs werecreated in the same country.

3.3.5. EstimationEstimation is usually carried out by taking the measured PE

content of a food similar to the target food. As this approach canonly yield rough approximations, estimated values should be re-placed as soon as analytical data become available. Still, it is con-sidered better to use rough estimations than to work with emptyfields as long as this is documented. Five of the reviewed PE DBscontain also estimated data and four of them indicate clearly whichof the data were estimated (Blitz et al., 2007; Park, Song, Joung, Li,& Paik, 2007; Ritchie et al., 2006; Valsta et al., 2003).

Estimation of PE contents on the basis of the PE to fibre ratio hasbeen described in several articles (Blitz et al., 2007; de Kleijn et al.,2001; Valsta et al., 2003) and is carried out mainly for multiple-ingredient foods having some PE containing ingredients in un-known ratios. However, this approach requires knowledge of thefibre content of both the product and the ingredients, and relieson the correctness of the PE content of the fibre rich ingredientsand on the assumption that ingredients poor in fibre are devoidof PEs.

3.4. Mode of data expression

Data can be expressed in native forms or as aglycone equiva-lents on a fresh weight basis (as-is, as consumed) or on a dryweight basis and as single values or after aggregation. The biolog-ically active compounds are the aglycones and concentrations ofnative forms are not always available. However, aglycone equiva-lents can easily be calculated from the contents of native formspublished in the literature so that it is advisable to present datain aglycone equivalents in a DB unless information on the nativeforms is required. Indeed, only comprehensive DBs for use by foodregulatory authorities give contents in native forms, too (Gry et al.,2007; Scalbert, 2007).

Usually, DBs present values on an as-is basis in order to facili-tate intake assessment. Early literature compendia and compre-hensive DBs enabling comparison of values on a different weightbasis present values (also) on a dry weight basis. In addition, de

H. Schwartz et al. / Food Chemistry 113 (2009) 736–747 743

Kleijn and co-workers gave contents only on a dry weight basis.Pillow and co-workers (1999) and Fink et al. (2006) did not statethe weight basis but comparison with the original references re-vealed that most values in the DB by Pillow et al. (1999) are givenon a dry weight basis.

Aggregation can improve the quality of data which are not rep-resentative due to lacking sampling plan. However, it is importantto keep the original data in the DB. If non-aggregated data are pre-sented and sufficient additional information about the sample isgiven (i.e. country of origin, year, and season), the users can selectthe best-suited data for their specific purposes. If, on the otherhand, none of the values in the DB are particularly suited for intakeassessment in one specific country or population group, it is moreadvisable to aggregate values for better representativeness. Aggre-gated data are presented in five DBs, both aggregated and non-aggregated data in two DBs and non-aggregated data in ten DBs(see Table 1).

3.5. Amount of additional information

The information that should be available for the DB user de-pends on the aim of the DB. DBs for expert users should containall relevant information visible for the users. Of greatest impor-tance are the reference to the data source, sample specific informa-tion as, for example, the part of the sample analysed (whole samplevs. edible portion, roots or leaves), any processing used, the coun-try of origin of the samples, the sampling year, the data source, thesampling plan, method of sample preparation, analysis, identifica-tion and quantitation in the case of analytical data, the mode ofdata expression and, if determined, the quality of the individualDB entries. DBs for intake assessment need not present all informa-tion on the user interface. However, origin, sampling year, numberof primary sample units making up the sample, mode of dataexpression and information on aggregation and data quality areimportant for selecting the best-suited values for intake assess-ment. In DBs which will regularly be updated, all relevant availableinformation about the data should be collected (even if not visibleto the end user) in order to make (future) comparisons, datareplacements and quality assessment possible and to assure trace-ability of the data.

The amount of additional information is highly variable for theindividual DBs (see Table 1). The DB containing most informationon the food samples and on the analytical methods used to gener-ate the data is the EuroFIR BASIS DB, followed by the USDA DB andthe French DB on polyphenols (provided these details are given inthe original articles). Further DBs providing detailed informationare the literature compendium by Reinli and Block (1996), the Ve-nus DB by Kiely et al. (2003), the DBs by Ritchie and co-workers(2006) and by Park et al. (2007), and the DBs by Milder et al.(2005) and Thompson et al. (2006), which contain solely analyticaldata. The least additional information is given in DBs or FCTs for in-take assessment by means of a FFQ.

3.6. Quality control of the data

It is essential to check that data meet minimum quality require-ments for unequivocal food description, data expression, and com-ponent identification before they are included into a DB. For eightof the 13 DBs using peer-reviewed literature as one of the datasources, it was stated that conflicting or low quality data werenot included into the DB. Of these eight DBs, five performed a qual-ity assessment prior to the inclusion of data (Kiely et al., 2003; Parket al., 2007; Ritchie et al., 2006; Scalbert, 2007; US Department ofAgriculture, 2002). Quality assessed data are given in five DBs (Finket al., 2006; Gry et al., 2007; Park et al., 2007; US Department ofAgriculture, 2002; Valsta et al., 2003).

None of the literature compendia contained quality rated data.However, Reinli and Block (1996) excluded equivocal data prior toinclusion into the DB, and the data listed in Mazur’s data collectionwere obtained by a validated analytical method. Of the four DBscreated for intake assessment by means of a FFQ, only the most re-cent one contained quality assessed data (Fink et al., 2006). Qualityassessment was performed according to the USDA system (seebelow).

In the seven DBs for intake assessment in different countriesor populations, quality assessment was performed in differentways: Kiely et al. (2003) evaluated the quality of papers providingPE contents prior to selection for inclusion into the DB accordingto five criteria: the sampling procedure, number of samples han-dled, sample handling, analytical method and analytical qualitycontrol. Of the 47 references studied, 14 were excluded due toinadequate analytical method, poorly described quantitation orunclear expression of results. After establishment of the DB, eachcalculation was checked independently (conversion from glyco-side into aglycone values, dry weight into wet weight and thestandardisation to lg/100 g). In addition, a random check of thedata was performed by a Venus partner and each data entrywas checked. However, data were not given an overall quality in-dex. In addition, no information about sampling was available inthe DB so that it is difficult to determine how representative dataare in the DB.

Ritchie and co-workers (2006) carried out quality control at twolevels: prior to inclusion of values into the DB and after establish-ment of the DB: for the data to be included into the DB, the follow-ing accompanying information had to be presented in the paper:type of food, country of origin; McCance and Widdowson code, ifthe food was raw or cooked; dry or wet weight basis, type of inter-nal standard and method of analysis (including coefficient of vari-ation, multiple sample analysis and appropriate level of detection).The established DB was tested using the method of duplicate dietanalysis, in which the dietary isoflavone intake of vegetariansand omnivores was assessed. In addition, the correlation betweenestimated isoflavone intake using the DB and measured isoflavoneintake from duplicate diet analysis was determined. The establish-ment and use of a sampling plan were not among the qualityrequirements.

Milder et al. (2005) carried out analytical quality control andtook care that samples were representative by assembling compos-ite samples consisting of sub-samples from three different loca-tions or of three major brands. Thompson and co-workers (2006)who also included solely own analytical values in their DB also per-formed sampling according to a sampling plan but did not reportanalytical quality control. Blitz et al. (2007) did not report anyquality control measures either, except that the quality of analyti-cal values was judged by the reported recovery and precision of theanalyses. Park et al. (2007), on the other hand, included only soundanalytical data and well-documented data obtained by appropriatesampling plans. In addition, they performed quality assessment ofanalytical data according to the USDA expert quality evaluationsystem (US Department of Agriculture, 2002). However, only 18%of the data in the DB were extracted from the literature so thatthey could be assigned a USDA confidence code.

The most comprehensive of the DBs for intake assessment con-taining quality assessed data is the DB developed by Valsta and co-workers (2003), who used four criteria for the data quality assess-ment: the number of samples (including sampling plan), analyticalsample handling and documentation, documentation of the analyt-ical method and analytical quality control. For each criterion, 0–3quality indices were given. Data having at least nine quality indiceswere assigned a confidence code A. Therefore, even papers withonly one sample analysed and hence containing little representa-tive data might have obtained the highest quality code provided

744 H. Schwartz et al. / Food Chemistry 113 (2009) 736–747

that the analytical method, sample handling and quality controlwere appropriate and well documented.

Of the three comprehensive DBs for nutrition research, twodetermined the data quality according to expert quality assess-ment systems: the USDA expert data quality assessment systemcontained five criteria (Holden, Bhagwat, & Patterson, 2002): sam-pling plan, number of samples, sample handling, analytical method(evaluation of the analytical method and of the analyst using thismethod) and analytical quality control. Specific questions were de-fined for each category and rating points were assigned to each re-sponse. The maximum score was 20 for each category. Theseratings for the five categories were summed to yield a single nu-meric quality index (QI) for one component in a certain food froma certain data source. When acceptable data from different sourceswere combined to yield an aggregated value, this value was given aconfidence code, which was determined from all the individual QIof the combined data. Upon aggregation, the categories ‘‘samplingplan” and ‘‘number of samples” were reassessed due to the inclu-sion of data for sample units from different regions or countries,which was likely to increase the representativeness of the value.The ratings for ‘‘sample handling”, ‘‘analytical method” and ‘‘ana-lytical quality control” were averaged from the combined data.

In the original BASIS DB, data were given a quality ratingexpressing how well the data met five criteria including plantdescription, sample handling, component identification, and appro-priateness of the analytical method and sampling plan. If one of thecriteria was not met, the highest confidence code could not bereached and the evaluator had to make a note, in which respectthe data were deficient. This quality rating system was not aban-doned in the new EuroFIR BASIS DB. Although quality assessmentis carried out using an expert quality evaluation scheme, the origi-nal data quality evaluation system is still kept for comparisons. Thenew quality rating system includes questions on the plantfood description, processing, sampling plan, sample handling, com-pound identification, analytical method and on the analytical per-formance. Each criterion is rated with a figure of 1–5,corresponding to a score of 3–14 or 15 points. Summation of theindividual quality scores gives the overall quality rating (maximum100 points). In contrast to the USDA DB and to the DB constructedby Valsta et al. (2003), the individual quality ratings are visible inthe EuroFIR BASIS DB, enabling the user to see where even gooddata are deficient or why data are of low quality. This is of greatadvantage when several data deficient in one or the other aspectare to be aggregated to give one more representative mean valueof higher quality.

To sum up, whereas the exclusion of conflicting data is suffi-cient for DBs intended for snap-shots of the PE exposure of thestudy population, DBs which will be used over a longer period oftime and which will regularly be updated should contain qualityassessed data entries to facilitate future data replacements. DBsfor expert users should have the individual quality ratings visibleto the end user.

4. Selection of the best-suited databases for different purposes

PE contents summarised in literature compendia, DBs for usewith FFQs and DBs containing exclusively analytical data fromchemical analysis by the authors as well as PE contents in the Kor-ean nutrient DB described by Park et al. (2007) are available fromthe papers describing the construction of the individual DBs. Like-wise, the electronic DB developed by Ritchie and co-workers(2006) and the USDA DB are internet-based with free use. How-ever, the DBs by Valsta et al. (2003) and Blitz et al. (2007) are pro-prietary DBs and are only for internal use. The Venus DB (Kielyet al., 2003) and the original BASIS DB are no longer available to

the public but are still existing on CD-ROM. The EuroFIR BASISDB is currently available only to the EuroFIR Consortium and its fu-ture accessibility has to be agreed on. On the other hand, theFrench DB on polyphenols will be fully accessible to the publicafter compilation which is now complete.

The main applications and uses of PE DBs are intake assessment,nutrition research and facilitation of further compositional re-search by providing an overview of the contents and native formsof PEs in different foodstuffs. However, not all DBs are equally sui-ted for each purpose. In the following, the suitability of the individ-ual DBs for (1) intake assessment and (2) nutrition research,including provision of information on native forms and on the var-iability of PE contents in foodstuffs, will be discussed.

4.1. Suitability of the available DBs for intake assessment

The main prerequisites for DBs allowing intake assessment arehigh coverage of the foods commonly consumed by the study pop-ulation, inclusion of top sources of PEs even if these are minor con-stituents of the diet of the population under study and inclusion ofas many PEs as possible. Food description should be detailed en-ough to enable judgment if data are representative for the foodsin the country of interest. In addition, data should be quality as-sessed and the quality codes should be visible to the user to allowthe detection of unreliable data.

4.1.1. IsoflavonesThe highest coverage of foodstuffs is achieved by the DB con-

structed by Ritchie and co-workers (2006), which contains �960differently processed foods with total isoflavone contentsP0.05 mg/kg. Limitations are that gly, the third soy isoflavone, isnot included, and that data are not quality assessed. In addition, itis necessary to go back to the original reference in order to obtaininformation about the origin of the sample. Valuable alternativesor additional data sources are the USDA DB and the DBs by Thomp-son et al. (2006) and Park et al. (2007). These DBs contain�130 fooditems and cover all soy isoflavones. The origin of the samples is pro-vided in the DBs by Thompson et al. and Park and co-workers, andconfidence codes are given for USDA values and analytical data inthe Korean nutrient DB. The literature compendium by Reinli andBlock (1996) should also be considered as potential data sourcefor intake assessment. Although gly is not included, values for formoand bio A are available and the data are accompanied by a lot ofadditional information of relevance for intake assessment.

The DB created by Horn-Ross et al. (2000) contains values forfoods as consumed by the study population (118 women in Califor-nia’s San Francisco Bay Area) at the time of the study. However, foodsources and preparation techniques differ between populations, andno information on the proportion of raw to prepared or cooked sub-samples or on the pooled sub-samples in general is available in thepaper. Hence, the results obtained for the investigated food samplesmay not be applicable to intake assessment in other areas of the US orin other countries. The further DBs either cover a very small selectionof foods (de Kleijn et al., 2001; Fink et al., 2006), present values on adry weight basis (Mazur, 1998) or neglect to state the weight basis(Pillow et al., 1999). A further problem associated with the use ofDBs for intake assessment by means of FFQs is that ‘‘similar” foodsare grouped to one FFQ food group item (de Kleijn et al., 2001; Finket al., 2006). For instance, in the FFQ used by de Kleijn and co-work-ers, apricots, peaches and plums are grouped to one item althoughtheir total lignan contents differ by the factor 5 according to the datapublished by Milder et al. (2005).

4.1.2. LignansThe two DBs of choice for the intake assessment of lignans were

created by chemical analysis by the DB authors (Milder et al., 2005;

H. Schwartz et al. / Food Chemistry 113 (2009) 736–747 745

Thompson et al., 2006). Both cover�110 foods and provide data forseco, matai, larici and pino obtained by the analysis of compositesamples (obtained by purchasing multiple units of one sample indifferent outlets) by state of the art analytical methods. The collec-tion of own analysis data by Mazur (1998) provides some addi-tional data for the contents of seco and matai in basic plantfoods and beverages. However, these data were obtained by conve-nience sampling and are presented on a dry weight basis. Never-theless, many of the data produced by Mazur and co-workerswere included in several of the DBs described in this work (deKleijn et al., 2001; Fink et al., 2006; Gry et al., 2007; Kiely et al.,2003; Meagher & Beecher, 2000; Pillow et al., 1999; Valsta et al., 2003).

4.2. Suitability of the available DBs for nutrition research

Unfortunately, the DBs intended and by far most suitable fornutrition research are currently under construction (Gry et al.,2007; Scalbert, 2007) and/or being updated (Gry et al., 2007; USDepartment of Agriculture, 2002). However, the currently availableversion of the USDA DB is also highly suitable because it providesdetailed food description and information on the range of isoflav-one contents in foodstuffs of the same kind, on the number of sam-ples analysed to give the reported value and on the quality of thedata. Of the further available DBs, the literature compendia byReinli and Block (1996) and Meagher and Beecher (2000) can beused for nutrition research. Reinli and Block (1996) present non-aggregated data in a well-structured way, which facilitates thecomparison of values for foods of the same type obtained by thesame or a similar method (e.g. different varieties or batches of soy-beans analysed in one or different references). In addition, a lot ofadditional information (food description, number of samples, ana-lytical method) is presented. Meagher and Beecher (2000) summa-rised the published contents of the mammalian lignan precursorsseco and matai obtained by direct analysis (data mainly from pa-pers published by Mazur and co-workers (Mazur, 1998; Mazur,Duke, Wahala, Rasku, & Adlercreutz, 1998; Mazur, Wahala, Wang,& Adlercreutz, 1998; Mazur et al., 1996; Mazur, Wahala, Rasku,et al., 1998)) and compared them with the contents of the mamma-lian lignans enterodiol and enterolactone obtained by in vitro fer-mentation (Thompson, Robb, Serraino, & Cheung, 1991), which isof interest with respect to bioavailability. Pillow and co-workers(1999) reported values for mammalian lignans, too, but did notencourage comparison with the contents of seco and matai, be-cause contents for plant and mammalian lignans were given in dif-ferent foodstuffs.

DBs relying on calculation and estimation as one of the maindata sources (Park et al., 2007; Ritchie et al., 2006) are suitablefor nutrition research to only a limited extent. They can providean overview of the PE content in processed composite foods but of-fer no information about variability, native forms or bioavailability.Likewise, DBs constructed from analytical data (Horn-Ross et al.,2000; Milder et al., 2005; Thompson et al., 2006) and DBs for usewith FFQs (de Kleijn et al., 2001; Fink et al., 2006) provide littleinformation for nutrition research. However, as the DBs by Milderet al. and Thompson et al. are the only available DBs containingvalues for seco, matai, larici and pino, they have to be used to-gether with original research articles (Penalvo, Haajanen, et al.,2005; Penalvo et al., 2008; Schwartz & Sontag, 2006; Smedset al., 2007) until the EuroFIR BASIS DB and the French DB on poly-phenols become accessible.

5. Discussion

The potential beneficial health effects of PEs and the need to as-sess the PE exposure in different countries and different population

groups have led to the construction of several PE DBs in the lastdecade. These DBs differ in the time of establishment, in structureand aim and therefore also in comprehensiveness. Literature com-pendia and DBs for intake assessment by means of FFQs containonly a limited number of foods (<100), whereas four of the sevenDBs for intake assessment in different countries cover 800–6000basic and processed foodstuffs. The compounds covered rangefrom just total isoflavones to all PEs for which literature data areavailable. The only data source for literature compendia and fortwo of the three comprehensive DBs for nutrition research is scien-tific literature, whereas data sources in DBs for intake assessmentare multiple except for the three DBs containing just data obtainedby own laboratory analysis. Most DBs present PE contents in agly-cone equivalents and on an as-is basis. However, the Basis DB andthe French DB on polyphenols present data as they are given in theliterature and two literature compendia as well as one DB for in-take assessment based on a FFQ give contents on a dry weight ba-sis. In two DBs, the weight basis is not stated. Aggregated data arepresented in five, non-aggregated data in nine DBs and three DBspresent both aggregated and non-aggregated data. The amount ofadditional information is highly variable. The least information isprovided in DBs for intake assessment based on FFQs, whereasDBs for nutrition research are most comprehensive. Measures toassess the data quality are also variable, ranging from no qualitycontrol via quality control prior to inclusion of data to qualityassessment using expert evaluation systems.

Nevertheless, despite these differences there are some commonlimitations of PE DBs, even if different approaches have been madeto minimise them. The main difficulty results from the high vari-ability of PE contents in foodstuffs due to both foodborne factors(cultivar, environmental conditions, different processing methods,use of different formulations) and sample work-up and analysis bydifferent methods (different extraction and hydrolysis conditions,different standards). In Table 2, the contents of selected isoflavonesand lignans in four different foodstuffs as given in the individualDBs are listed.

The values differ most for the total lignan contents in carrotsand broccoli. Especially the high difference between data pub-lished by Milder et al. (2005) and data published by Thompsonet al. (2006) is remarkable. The values given by Thompson aremore in line with the dry weight data determined and given byMazur (1998) and overtaken by several other DBs. The fact thatone DBs presents wrong values (due to a comma error) stressesthe importance of cross checking the established DBs.

The contents of seco in flaxseed were similar in all DBs if the fig-ure 817 mg/kg (obtained by just enzymatic hydrolysis) was disre-garded. However, much higher values have already been publishedin the literature (up to 29,300 mg/kg secoisolariciresinol corre-sponding to 1540 mg/kg seco aglycone (Eliasson et al., 2003)) sothat intake assessment with the DB values may result in the under-estimation of the lignan intake due to insufficient coverage of thenatural variability of the content of seco in flaxseed in the availableDBs. DB contents of gen and total isoflavones in soybeans are betterin reflecting the natural variability. Contents of total isoflavoneaglucones ranging from 180 to 5620 mg/kg (Kim, Jung, Ahn, &Chung, 2005; Lee et al., 2003) have been determined in raw, ma-ture soybeans of different varieties grown in different locationsin the same year, and minimum and maximum DB values are373 and 2389 mg/kg, respectively.

There are approaches to minimise the impact of the natural var-iability and the variability introduced by processing both at thestage of data generation and at the stage of data presentation. Ana-lytical chemists can enhance the representativeness of their databy establishing and using a proper sampling plan covering variabil-ity in region, time and variety or brand instead of performing con-venience sampling at the nearest supermarket. DB managers have

Table 2Comparison of lignan and isoflavone contents of selected foods listed in different DBs

Reference Total lignans Seco Total isoflavones Gen

Carrot Broccoli Flax Soybeans Soybeans

Reinli and Block (1996) 735–2389 428–1382Mazur (1998) 1.95 (DW) 4.37 (DW) 3699 (DW) 373–1875 (DW) 268–1025 (DW)Meagher and Beecher (2000) 4.78 (DW) – comma error 4.37 (DW) 817–3699 (DW) – –Pillow et al. (1999) 1.95 (DW) 4.37 (DW) 2258 (DW?) 1646 (soy flour) 939 (soy flour)Horn-Ross et al. (2000) 0.38 Traces (< 0.25 mg/kg) Not in the DB 469 198de Kleijn et al. (2001) 3700 (DW) 4.37 (DW) Not in the DB Not in the DB Not in the DBFink et al. (2006) 3.5 (DW?) 4.4 (DW?) Not in the DB Not in the DB Not in the DBRitchie et al. (2006) – – – 1421 No values for individual

isoflavonesMilder et al. (2005) 1.71 13.25 2942 – –Thompson et al. (2006) 0.06 0.94 3753 1036 442Park et al. (2007) – – – 568–986 (dried yellow

soybeans)334–466 (dried yellowsoybeans)

US Department of Agriculture(2002)

– – – 1284 (362–2209) 738 (207–1341)

Values in mg/kg fresh weight unless stated otherwise.–: PE not covered in the DB.

746 H. Schwartz et al. / Food Chemistry 113 (2009) 736–747

the possibility to aggregate values for samples from different coun-tries, sampling years and varieties or brands. Dealing with lowquality analytical data is more difficult. Due to the limited amountof analytical data for PE contents in foods even low quality valuesobtained by outdated or not validated methods are included insome DBs. These values as well as values obtained by estimationare preferable to blank data fields in DBs for intake assessment.Still, it is imperative to give them a low quality code if the DB isregularly updated so that they can be replaced as soon as morereliable data become available.

Besides using a sampling plan and validated up to date samplepreparation and analytical methods and besides carrying out ana-lytical quality control, analytical chemists can considerably im-prove the quality of their data by accurately describing the foodsamples that are analysed (common and latin name, cultivar/brand, country and region of origin, year and season, maturity,manufactures, processing), the sampling plan (number of primarysample units, analysis of composite or individual samples), samplehandling, sample work-up and analysis (method of identification,method of quantitation) and analytical quality control. In addition,results should be given in aglycone equivalents on a wet weightbasis together with the moisture content to facilitate comparisonwith other data.

Even if the difficulties due to foodborne variability, analysis bydifferent methods and badly documented analytical data are min-imised, there are still problems likely to result in the underestima-tion of the PE intake in Western countries. Especially in the case ofliterature compendia and early DBs, the food coverage in the liter-ature was limited and some data could not be used because impor-tant information (e.g. the form of the foods being analysed) wasmissing (Pillow et al., 1999). One of the main problems then andnowadays is the possible use of soy ingredients in the productionof frequently consumed basic foods like bread, donuts, sausagesor certain canned foods which would not be recognised as PEsource in the first place (de Kleijn et al., 2001; Horn-Ross et al.,2000; Pillow et al., 1999). If laboratory analysis is not possible,the manufacturers of the main brands of these products shouldbe asked whether soy ingredients are used and, if yes, in whichproportions in order to enable estimation of the PE contents. A fur-ther problem resulting in the underestimation of the PE intake isthat, although syringa and 7-OH-matai have been identified asmain lignans in cereals and oil seeds, very few data on these andfurther, perhaps not even identified but possibly abundant lignansare currently available.

6. Conclusion

Despite the abovementioned limitations, PE DBs are valuablefor identifying PE sources, for providing a range of PE contents infoodstuffs, for intake assessment based on FFQs, dietary recallsand data from national food consumption studies and, in case suf-ficient additional information is provided, for nutrition research,product developers in the food industry and the regulatory affairssector. It must, however, be stressed that the existing PE DBs arenot intended and not suitable for use in human intervention stud-ies, where PE contents of the administered foods need to be exactlyknown and therefore have to be determined by chemical analysis.Nevertheless, due to ongoing research especially in lignan analysis,a greater amount of foodstuffs will be covered in the future andeventually several data entries will be available for individualfoodstuffs. Hence, the chances will be higher to find data for foodsconsumed by a certain population group or in a certain country. Ifone certain food of interest is not covered (e.g. many data for iso-flavone contents of tofu from Germany, Switzerland and Italy butno entries for tofu from Austria), aggregation of the available val-ues will yield more representative data and intake assessment willbecome more reliable and less dependent on one individual dataentry. In order to make use of the future analytical values, it isimperative that DBs are regularly updated and that quality codesare given showing which data are in which way deficient and needto be replaced. Therefore, although a lot of effort has already beenmade in the area of PE DBs, plenty of work will still be required toassure the sustainability of the existing DBs, to enhance the repre-sentativeness of the data and to increase the number of foods cov-ered by regular updates.

Acknowledgements

This work was completed on behalf of the EuroFIR Consortium(FOOD-CT-2005-513944) and funded under the EU 6th FrameworkFood Quality and Safety Programme.

References

Achouri, A., Boye, J. I., & Belanger, D. (2005). Soybean isoflavones: Efficacy ofextraction conditions and effect of food type on extractability. Food ResearchInternational, 38, 1199–1204.

Adlercreutz, H. (2007). Lignans and human health. Critical Reviews in ClinicalLaboratory Science, 44, 483–525.

Blitz, C. L., Murphy, S. P., & Au, D. L. M. (2007). Adding lignan values to a foodcomposition database. Journal of Food Composition and Analysis, 20, 99–105.

H. Schwartz et al. / Food Chemistry 113 (2009) 736–747 747

Cassidy, A., Albertazzi, P., Nielsen, I. L., Hall, W., Williamson, G., Tetens, I., et al.(2006). Critical review of health effects of soybean phyto-oestrogens in post-menopausal women. Proceedings of the Nutrition Society, 65, 76–92.

Collison, M. W. (2008). Determination of total soy isoflavones in dietarysupplements, supplement ingredients, and soy foods by high-performanceliquid chromatography with ultraviolet detection: Collaborative study. Journalof AOAC International, 91, 489–500.

Coward, L., Smith, M., Kirk, M., & Barnes, S. (1998). Chemical modification ofisoflavones in soyfoods during cooking and processing. American Journal ofClinical Nutrition, 68, 1486S–1491S.

de Kleijn, M. J., Van der Schouw, Y. T., Wilson, P. W. F., Adlercreutz, H., Mazur, W.,Grobbee, D. E., et al. (2001). Intake of dietary phytoestrogens is low inpostmenopausal women in the United States: The Framingham Study. Journal ofNutrition, 131, 1826–1832.

Eliasson, C., Kamal-Eldin, A., Andersson, R., & Aman, P. (2003). High-performanceliquid chromatographic analysis of secoisolariciresinol diglucoside andhydroxycinnamic acid glucosides in flaxseed by alkaline extraction. Journal ofChromatography A, 1012, 151–159.

Erdman, J. W., Jr., Badger, T. M., Lampe, J. W., Setchell, K. D., & Messina, M. (2004).Not all soy products are created equal: Caution needed in interpretation ofresearch results. Journal of Nutrition, 134, 1229S–1233S.

Fink, B. N., Steck, S. E., Wolff, M. S., Kabat, G. C., & Gammon, M. D. (2006).Construction of a flavonoid database for assessing intake in a population-basedsample of women on Long Island, New York. Nutrition and Cancer, 56, 57–66.

Griffith, A. P., & Collison, M. W. (2001). Improved methods for the extraction andanalysis of isoflavones from soy-containing foods and nutritional supplementsby reversed-phase high-performance liquid chromatography and liquidchromatography–mass spectrometry. Journal of Chromatography A, 913,397–413.

Gry, J., Black, L., Eriksen, F., Pilegaard, K., Plumb, J., Rhodes, M., et al. (2007). EuroFIRBASIS database. Trends in Food Science & Technology, 18, 434–444.

Holden, J., Bhagwat, S. A., & Patterson, K. Y. (2002). Development of a multi-nutrientdata quality evaluation system. Journal of Food Composition and Analysis, 15,339–348.

Horn-Ross, P. L., Barnes, S., Lee, M., Coward, L., Mandel, J. E., Koo, J., et al. (2000).Assessing phytoestrogen exposure in epidemiologic studies: Development of adatabase (United States). Cancer Causes Control, 11, 289–298.

Johnson, P., Kamal-Eldin, A., Lundgren, L. N., & Aman, P. (2000). Method for analysisof secoisolariciresinol diglucoside in faxseeds. Journal of Agricultural and FoodChemistry, 48, 5216–5219.

Kamal-Eldin, A., Peerlkamp, N., Johnsson, P., Andersson, R., Andersson, R. E.,Lundgren, L. N., et al. (2001). An oligomer from flaxseed composed ofsecoisolariciresinoldiglucoside and 3-hydroxy-3-methyl glutaric acid residues.Phytochemistry, 58, 587–590.

Kiely, M., Faughnan, M., Wahala, K., Brants, H., & Mulligan, A. (2003). Phyto-oestrogen levels in foods: The design and construction of the VENUS database.British Journal of Nutrition, 89(Suppl 1), S19–S23.

Kim, S. H., Jung, W. S., Ahn, J. K., & Chung, I. M. (2005). Analysis of isoflavoneconcentration and composition in soybean [Glycine max (L.)] seeds between thecropping year and storage for 3 years. European Food Research and Technology,220, 207–214.

Kraushofer, T., & Sontag, G. (2002). Determination of some phenolic compounds inflax seed and nettle roots by HPLC with coulometric electrode array detection.European Food Research and Technology, 215, 529–533.

Lee, S. J., Ahn, J. K., Kim, S. H., Kim, J. T., Han, S. J., Jung, M. Y., et al. (2003). Variationin isoflavone of soybean cultivars with location and storage duration. Journal ofAgricultural and Food Chemistry, 51, 3382–3389.

Liggins, J., Bluck, L. J., Coward, W. A., & Bingham, S. A. (1998). Extraction andquantification of daidzein and genistein in food. Analytical Biochemistry, 264,1–7.

Liggins, J., Grimwood, R., & Bingham, S. A. (2000). Extraction and quantification oflignan phytoestrogens in food and human samples. Analytical Biochemistry, 287,102–109.

Mazur, W. (1998). Phytoestrogen content in foods. Baillieres Clinical Endocrinologyand Metabolism, 12, 729–742.

Mazur, W., & Adlercreutz, H. (1998). Naturally occurring oestrogens in food. Journalof Pure and Applied Chemistry, 70, 1759–1776.

Mazur, W. M., Duke, J. A., Wahala, K., Rasku, S., & Adlercreutz, H. (1998).Isoflavonoids and lignans in legumes: Nutritional and health aspects inhumans. Journal of Nutritional Biochemistry, 9, 193–200.

Mazur, W., Fotsis, T., Wahala, K., Ojala, S., Salakka, A., & Adlercreutz, H. (1996).Isotope dilution gas chromatographic–mass spectrometric method for thedetermination of isoflavonoids, coumestrol, and lignans in food samples.Analytical Biochemistry, 233, 169–180.

Mazur, W. M., Wahala, K., Rasku, S., Salakka, A., Hase, T., & Adlercreutz, H. (1998).Lignan and isoflavonoid concentrations in tea and coffee. British Journal ofNutrition, 79, 37–45.

Mazur, W., Wahala, K., Wang, G., & Adlercreutz, H. (1998). Dietary phytoestrogens –From chemistry to chemoprevention. Kemia-Kemi, 25, 48–55.

Meagher, L. P., & Beecher, G. R. (2000). Assessment of data on the lignan content offoods. Journal of Food Composition and Analysis, 13, 935–947.

Messina, M., Kucuk, O., & Lampe, J. W. (2006). An overview of the health effects ofisoflavones with an emphasis on prostate cancer risk and prostate-specificantigen levels. Journal of AOAC International, 89, 1121–1134.

Milder, I. E., Arts, I. C. W., van de Putte, B., Venema, D. P., & Hollman, P. C. H. (2005).Lignan contents of Dutch plant foods: A database including lariciresinol,

pinoresinol, secoisolariciresinol and matairesinol. British Journal of Nutrition, 93,393–402.

Mitjans, M., & Vinardell, M. P. (2005). Biological activity and health benefits oflignans and lignins. Trends in Comparative Biochemistry & Physiology, 51,857–866.

Moazzami, A. A., & Kamal-Eldin, A. (2006). Sesame seed is a rich source of dietarylignans. JAOCS, 83, 719–723.

Mortensen, A., Kulling, S. E., Schwartz, H., Rowland, I., Ruefer, C., Rimbach, G., et al.(2009). Analytical and compositional aspects of isoflavones in food and theirbiological effects. Journal of Molecular Nutrition and Food Research, in press.

Muellner, C., & Sontag, G. (1999). Determination of some phytoestrogens insoybeans and their processed products with HPLC and coulometric electrodearray detection. Fresenius Journal of Analytical Chemistry, 364, 261–265.

Obermeyer, W. R., Musser, S. M., Betz, J. M., Casey, R. E., Pohland, A. E., & Page, S. W.(1995). Chemical studies of phytoestrogens and related compounds in dietarysupplements: Flax and chaparral. Proceedings of the Society for ExperimentalBiology and Medicine, 208, 6–12.

Official Methods of Analysis (2000). AOAC Official Method 2001.10: Determinationof isoflavones in soy and selected foods containing soy. Chapter 45.44.14.

Park, M. K., Song, Y., Joung, H., Li, S. J., & Paik, H. Y. (2007). Establishment of anisoflavone database for usual Korean foods and evaluation of isoflavone intakeamong Korean children. Asia Pacific Journal of Clinical Nutrition, 16, 129–139.

Penalvo, J. L., Adlercreutz, H., Uehara, M., Ristimaki, A., & Watanabe, S. (2008).Lignan content of selected foods from Japan. Journal of Agricultural and FoodChemistry, 56, 401–409.

Penalvo, J. L., Haajanen, K. M., Botting, N., & Adlercreutz, H. (2005). Quantification oflignans in food using isotope dilution gas chromatography/mass spectrometry.Journal of Agricultural and Food Chemistry, 53, 9342–9347.

Penalvo, J. L., Heinonen, S.-M., Aura, A.-M., & Adlercreutz, H. (2005). Dietary sesaminis converted to enterolactone in humans. Journal of Nutrition, 135, 1056–1062.

Penalvo, J. L., & Nurmi, T. (2006). Application of coulometric electrode arraydetection to the analysis of isoflavonoids and lignans. Journal of Pharmaceuticaland Biomedical Analysis, 41, 1497–1507.

Penalvo, J. L., Nurmi, T., & Adlercreutz, H. (2004). A simplified HPLC method for totalisoflavones in soy products. Food Chemistry, 87, 297–305.

Pillow, P. C., Duphorne, C. M., Chang, S., Contois, J. H., Strom, S. S., Spitz, M. R., et al.(1999). Development of a database for assessing dietary phytoestrogen intake.Nutrition and Cancer, 33, 3–19.

Reinli, K., & Block, G. (1996). Phytoestrogen content of foods – A compendium ofliterature values. Nutrition and Cancer, 26, 123–148.

Ritchie, M. R., Cummings, J. H., Morton, M. S., Michael Steel, C., Bolton-Smith, C., &Riches, A. C. (2006). A newly constructed and validated isoflavone database forthe assessment of total genistein and daidzein intake. British Journal of Nutrition,95, 204–213.

Ritchie, M. R. (n.d.). Phyto-oestrogen database. <http://medicine.st-andrews.ac.uk/research/docs/ritchie/> Accessed 08.09.08.

Sacks, F. M., Lichtenstein, A., Van Horn, L., Harris, W., Kris-Etherton, P., & Winston,M. (2006). Soy protein, isoflavones, and cardiovascular health: A summary of astatement for professionals from the American Heart Association NutritionCommittee. Arteriosclerosis Thrombosis and Vascular Biology, 26, 1689–1692.

Scalbert, A. (2007). French database on polyphenols, personal communication.Schwartz, H. (2008). Analysis of bioactive compounds in foods: Isoflavones, tocopherols,

tocotrienols and plant sterols. Ph.D. Thesis. University of Vienna. <http://www.univie.ac.at/anchem/dep2/PhD-thesis%20Schwartz.pdf> Accessed08.09.08.

Schwartz, H., & Sontag, G. (2006). Determination of secoisolariciresinol, lariciresinoland isolariciresinol in plant foods by high performance liquid chromatographycoupled with coulometric electrode array detection. Journal of ChromatographyB, 838, 78–85.

Setchell, K. D., Brown, N. M., Desai, P., Zimmer-Nechemias, L., Wolfe, B. E., Brashear,W. T., et al. (2001). Bioavailability of pure isoflavones in healthy humans andanalysis of commercial soy isoflavone supplements. Journal of Nutrition, 131,1362S–1375S.

Setchell, K. D. R., Childress, C., Zimmer-Nechemias, L., & Cai, J. (1999). Method formeasurement of dietary secoisolariciresinol using HPLC with multichannelelectrochemical detection. Journal of Medicinal Food, 2, 193–198.

Setchell, K. D., & Cole, S. J. (2003). Variations in isoflavone levels in soy foods and soyprotein isolates and issues related to isoflavone databases and food labeling.Journal of Agricultural and Food Chemistry, 51, 4146–4155.

Smeds, A. I., Eklund, P. C., Sjoholm, R. E., Willfor, S. M., Nishibe, S., Deyama, T., et al.(2007). Quantification of a broad spectrum of lignans in cereals, oilseeds, andnuts. Journal of Agricultural and Food Chemistry, 55, 1337–1346.

Thompson, L. U., Boucher, B. A., Liu, Z., Cotterchio, M., & Kreiger, N. (2006).Phytoestrogen content of foods consumed in Canada, including isoflavones,lignans, and coumestan. Nutrition and Cancer, 54, 184–201.

Thompson, L. U., Robb, P., Serraino, M., & Cheung, F. (1991). Mammalian lignanproduction from various foods. Nutrition and Cancer, 16, 43–52.

US Department of Agriculture, A. R. S. (2002). USDA – Iowa State Universitydatabase on the isoflavone content of foods. Release 1.3, Nutrient DataLaboratory. <http://www.nal.usda.gov/fnic/foodcomp/Data/isoflav/isoflav.html>Accessed 08.09.08.

Valsta, L. M., Kilkkinen, A., Mazur, W., Nurmi, T., Lampi, A. M., Ovaskainen, M. L.,et al. (2003). Phyto-oestrogen database of foods and average intake in Finland.British Journal of Nutrition, 89(Suppl 1), S31–38.

Westcott, N. D., & Muir, A. D. (2003). Flax seed lignan in disease prevention andhealth promotion. Phytochemistry Reviews, 2, 401–417.