SEARCH

SEARCH BY CITATION

Keywords:

  • Database;
  • dietary assessment;
  • dietary carbohydrates;
  • glycaemic index

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of Interest Statement
  8. Acknowledgement
  9. References

There is growing evidence that the glycaemic index (GI) of the diet is important with respect to body weight and metabolic disease risk. However, research is limited by the paucity of GI values for commonly consumed carbohydrate-rich foods in European countries. A new methodology has been developed for consistent assignment of GI values to foods across five European databases used in the Diogenes intervention study. GI values were assigned according to five decreasing levels of confidence (1) Measured values for specific foods; (2) Published values from published sources; (3) Equivalent values where published values for similar foods existed; (4) Estimated values assigned as one of three values representing low/medium/high GI ranges and (5) Nominal values assigned as 70, where no other value could be assigned with sufficient confidence. GI values were assigned to 5105 foods. In food records collected at baseline, the contribution to carbohydrate intake of foods assigned levels 1–2 ranged from 16% to 43% depending on country, and this increased to 53–81% including level 3 foods. The degree of confidence to assigned GI values differed across Europe. This standardized approach of assigning GI values will be made available to other researchers to facilitate further investigation into the effects of dietary GI on health.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of Interest Statement
  8. Acknowledgement
  9. References

There is growing evidence that the type of carbohydrate in the diet is important with respect to body weight and metabolic disease (1). The glycaemic index (GI), which classifies carbohydrate-containing foods based on their blood glucose-raising potential (2), has been the focus of significant interest from scientists, health professionals and the public, with respect to its impact on health.

However, there are barriers to the use of the GI concept in dietary intervention studies. First, in designing intervention studies to investigate the health effects of modifications to dietary GI, it is imperative that a difference in dietary GI is achieved. In order to accurately select high and low GI diets, the GI values of component foods must be known. The bulk of the published GI values arise from Australia and North America, and there is a paucity of measured GI values for commonly consumed European foods.

Second, the GI of a food is a physiological measure of the relative glycaemic response between foods, rather than a standard chemical assay, as for most nutrients. GI values are not included in any national food composition databases. This means that values are not linked to the nutrient information of the foods, including the carbohydrate content, which is required for the determination of both dietary GI and the glycaemic load (GL). Dietary GI is calculated as the weighted average of the GI values of all foods consumed (3). The GL reflects both the GI of a food and the quantity consumed and is calculated as the amount of carbohydrate in a food multiplied by the GI of the food and the quantity consumed, summed for all foods eaten (4). In order that dietary GI can be calculated from reported food intake records alongside other dietary variables, GI values need to be incorporated into nutrient databases. As measured GI values do not exist for many foods, assumptions must be made in assigning values. The majority of published studies that have calculated dietary GI from food intake records do not detail how values were assigned or the assumptions made, and the level of confidence in resulting dietary GI estimates is unclear. Several different approaches to this have recently been described in the literature (5–7), but there is no published standardized method for use in national food databases. For transparency and to improve consistency, such databases need to be publicly available. This would also save much repetition of effort and make more efficient use of resources.

Diogenes (Diet, Obesity and Genes) is a pan-European multicentre research study into the obesity epidemic (http://www.diogenes-eu.org). As part of this project, a dietary intervention study has been carried out in eight European centres in Maastricht (the Netherlands), Copenhagen (Denmark), Cambridge (UK), Heraklion (Greece), Pamplona (Spain), Potsdam (Germany), Sofia (Bulgaria) and Prague (the Czech Republic). The objective of the dietary intervention was to investigate the efficacy of diets varying in GI and protein content for weight control in overweight/obese individuals and their children (8). Food diaries were used to assess whether the study participants achieved dietary targets.

In this and other multicentre studies, where food diaries are coded using separate local food composition databases, it is critical that a consistent approach is taken across participating countries for the assignment of GI values to foods. This paper describes a method for assigning GI values to foods, which provides the foundations for a European GI database.

Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of Interest Statement
  8. Acknowledgement
  9. References

The protocol and dietary intervention strategy for the Diogenes study have been described in detail elsewhere (8,9).

Assessment of dietary intake

Dietary intake was assessed by weighed 3-d food diaries at baseline when subjects were consuming their habitual diet, at 2–4 weeks into the intervention, and after 26 weeks in all centres. In two centres (Denmark and the Netherlands), dietary intake was also assessed after 52 weeks. A standardized diary was used in all centres, translated into local languages (9).

Food diaries were coded locally in each centre and national food databases were used where available (UK (10), Denmark (11), the Netherlands (12), Germany (13), the Czech Republic (14)). The UK McCance and Widdowson database (10) was used by Spain, Bulgaria and Greece. Foods reported to be consumed but missing from databases were added, together with nutrient values provided by the manufacturer. If no such information was available a food in the database closely matching the description of the consumed food was coded. GI values were assigned to foods according to the procedure described below, and added into food databases as an additional variable.

Available carbohydrate was described as monosaccharide equivalents (MSE) in the UK database and was therefore slightly higher compared with other databases. To adjust for this difference the carbohydrate values from the UK database were standardized using the formula below (6), where starch, monosaccharide and disaccharide values were all available in the food composition database (10):

  • image

Once food diaries had been coded locally, the information was entered into a central database. The nutrient and GI intake was then calculated centrally to ensure consistency across the study centres. Before calculating the nutrient intake food weights and codes were checked and implausible values were checked in the food diaries and corrected where appropriate.

Assignment of glycaemic index values

The GI values relative to a glucose solution standard (3) were assigned to all foods appearing in food diaries and containing >0.1 g total carbohydrate per 100 g. Assignment was performed according to five levels of confidence, as follows:

Confidence level 1: measured values

Where the GI of a specific food had been measured at one of the research centres or by the manufacturer according to the published standardized protocol (15), this value was assigned with the highest level of confidence. GI values of some of the foods used in the intervention had previously been measured at MRC Human Nutrition Research (16), Maastricht University (van Baak MA, unpublished data) and the University of Copenhagen (Sloth B, unpublished data).

Confidence level 2: published values

If the GI value of a food had not been specifically measured, but a published value existed for a food of the same description, this was assigned. Various published sources were searched (17–20). Values were selected if the item description and preparation/cooking method (if given) matched. Where more than one GI value was found matching the description, an average value was calculated and used. In cases where the brand of the item matched, this value was selected in preference to others.

Confidence level 3: equivalent values

Where no published value exactly matched the item description, but a closely equivalent food was identified, this value was used. This was based on local knowledge of the ingredients, preparation and cooking method of an item, and a value was used as an equivalent if these were all comparable between foods. For example, a published value for beef and ale casserole of 53 was assigned to other slow-cooked meat dishes with similarly described sauces (18).

Confidence level 4: estimated values

Where no published value existed for a food item of sufficiently similar description, an estimate was made to one of three GI values: 45, 63 or 85, selected to represent low, medium and high GI ranges based on the midpoint of each category (21). Estimates were made as to which was most likely to be closest to the true value, based on the constituent ingredients and the processing and preparation methods. Rules were developed for certain types of food and dishes. These are described below and summarized in Table 1.

Table 1.  Estimates for certain food groups
EstimateFood
Low (45)Pulses without published values
Pulse-based dishes
Meat casseroles with added pulses
Pasta-based dishes
Non-starchy vegetables without published values
Non-starchy vegetable-based dishes
Medium (63)Non-specified rice
Non-specified rice-based dishes
Meat or vegetable casseroles with added potatoes
Starchy vegetables without published values
Pastry-based dishes
High (85)Non-specified potatoes
Potatoes without published values
Potato-based dishes

All non-starchy vegetables and mixed vegetable dishes comprising only non-starchy vegetables with no known GI value were assigned a low estimate of 45. This was based on the calculation by Flood et al. that the mean GI of all measured non-starchy vegetables was 32 (5). The value 45 is slightly higher than 32, but was used for consistency.

All mixed dishes comprising predominantly pasta or pulses were also assigned low estimates as all published values for pastas and pulses are low. Although the GI of potatoes varies somewhat, all have relatively high values. Potatoes or dishes consisting predominantly of potato were therefore assigned a high estimate of 85 where no published value existed.

Addition of significant quantities of carbohydrate-rich staple foods to mixed dishes was also taken into account. For example, meat casseroles were assigned the value of 53 (see above). However, if a casserole contained meat and pulses (all known to be low GI) the item was assigned the low estimate of 45. The pulses, having a lower GI than 53, would reduce the overall GI of the dish. If the casserole contained potatoes (which all have relatively high GI values) in addition to the meat, the item was assigned the medium estimate of 63. The potatoes, having a higher GI than 53, would raise the overall GI of the dish. These assumptions were made based on previous evidence that the GI of a meal can be predicted from its constituent components (22).

Rice and rice dishes were more difficult to estimate. There is wide variation in GI between different types of rice, with published values ranging from 27 for parboiled Bangladeshi high-amylose rice to 109 for Jasmine rice (20). However, as the majority fall into the medium GI category where the type of rice was not specified (either alone or as part of a rice-based dish), a medium estimate was assigned.

The purpose of assigning these estimate values was to avoid the excessive over- or under-estimation of dietary GI, which would result from assigning any single value to all of these foods.

Confidence level 5: nominal values

Where an estimate could not be made with sufficient confidence due to lack of information about the item or published values for similar or constituent foods, a nominal value of 70 was assigned. This is in line with methodology used previously by other researchers (23). The purpose of this was to avoid biasing the overall calculated dietary GI through missing values being assigned zero.

Flour was assigned the nominal value. Flour is occasionally reported in food diaries as a recipe component; however, as it is inedible uncooked, it is not possible to measure the GI value.

Procedure and coordination of glycaemic index assignment

Once these rules for assignment had been developed, the UK and Netherlands centres began assigning GI values to their databases. These values were then sent to Denmark, Germany and the Czech Republic, together with detailed instructions for assignment. These centres assigned GI values to foods in their databases, ensuring all were translated into English. Once completed, these databases were sent to the UK and Netherlands for checking that values had been assigned according to the rules and cross-checking that values for similar foods had been assigned consistently across all databases.

Data analysis

Dietary GI was calculated as the weighted average of the GI values of the foods consumed, according to the following equation (3):

  • image

The distribution of foods assigned GI values across differing levels of confidence was examined and compared between databases, and stratified according to carbohydrate content of the mean servings of each food. The proportion of reported dietary carbohydrate intake from each GI confidence level was calculated by centre, in order to make inferences about the confidence in the dietary GI results in each centre. This was calculated separately for baseline (habitual) intakes and during intervention, as the intervention included advice to change the GI of the diet. The effect of the selection of 70 as the nominal confidence level 5 value was then examined by replacing this with the mean dietary GI and with zero.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of Interest Statement
  8. Acknowledgement
  9. References

The total number of foods coded from diaries was 6058. Of these, 953 contained ≤0.1 g carbohydrate per 100 g, and were not assigned values. GI values were assigned to all 5105 remaining foods. The number and distribution of these foods across categories of GI assignment and by carbohydrate servings is shown in Table 2. These data are given overall and also stratified by database.

Table 2.  Number and % of total number of assigned carbohydrate-containing foods (n = 5105) by level of confidence, contribution of carbohydrate per mean serving and food composition table
Glycaemic index confidence levelDatabaseAll foodsCarbohydrate content (g per mean serving of the food)
<1010–2020–30>30
N (%) N (%) N (%) N (%) N (%)
  • * 

    M&W, McCance and Widdowson food composition tables (10). Used by the UK, Greece, Spain and Bulgaria.

1: measuredAll51 (1.0)1 (<0.1)6 (0.1)17 (0.3)27 (0.5)
M&W*2005510
Danish1900613
Dutch121164
Czech00000
German00000
2: publishedAll925 (18.1)251 (4.9)245 (4.8)164 (3.2)265 (5.2)
M&W*41510511175124
Danish20540513282
Dutch13945432328
Czech5519121113
German11142282318
3: equivalentAll1443 (28.3)509 (10.0)259 (5.1)211 (4.1)464 (9.1)
M&W*650183125113229
Danish319128514199
Dutch15271322227
Czech1054618833
German21780332876
4: estimatedAll1479 (29.0)647 (12.7)257 (5.0)227 (4.4)348 (6.8)
M&W*59522597112161
Danish204105352341
Dutch249144383037
Czech16457342647
German267116533662
5: nominalAll1207 (23.6)883 (17.3)159 (3.1)71 (1.4)94 (1.8)
M&W*417287623236
Danish288222351318
Dutch21216327715
Czech1158015713
German175131201212

Overall, one-fifth (19.1%) of the foods were assigned GI values at confidence levels 1 or 2 (i.e. the value had been specifically measured or published in the literature). Including foods assigned confidence level 3 values (i.e. foods for which a published value for at least a close-equivalent existed) increased this figure to about half of all foods (47.4%).

An additional 23.6% of the foods were assigned the confidence level 5 nominal value of 70, meaning that there was no data available on which to base any estimate of GI with sufficient confidence. Examination of the distribution of these confidence level 5 foods across carbohydrate content per mean serving revealed that 73% of the confidence level 5 foods provided less than 10 g per serving. Just 94 foods assigned confidence level 5 provided more than 30 g per serving. Typically, these foods were infrequently consumed, 61% were recorded at less than three eating occasions.

At baseline, before dietary intervention, the proportion of total energy intake from carbohydrate differed between study centres, ranging from 36.8% in Bulgaria to 45.9% in the Czech Republic (see Table 3). The contribution of different food groups to carbohydrate intake also varied, but bread was the largest contributor in all centres, providing from 16% of dietary carbohydrate in Greece to 32% in the Netherlands.

Table 3.  Carbohydrate intake as % of total energy intake at baseline
CentreMean (SD) carbohydrate intake (% total energy intake)
Bulgaria36.8 (8.7)
Czech Republic45.9 (9.5)
Denmark45.5 (7.6)
Germany42.8 (7.5)
Greece39.3 (7.2)
Netherlands43.2 (6.3)
Spain37.6 (8.6)
UK44.9 (7.1)

The proportion of dietary carbohydrate intake from foods in each GI confidence level was calculated. For reported habitual intakes (at baseline, before intervention), this is shown in Fig. 1. The contribution to carbohydrate intake of foods assigned confidence levels 1 or 2 ranged from 15.5% (the Czech Republic) to 42.8% (the Netherlands). Including foods assigned confidence level 3 raised the proportion of carbohydrate intake for which a GI value was available to between 53.2% (the Czech Republic) and 81.0% (Denmark). Conversely, the proportion of carbohydrate intake provided by confidence level 5 foods, for which no GI value was available, ranged from 5% to 6% (Denmark, the Netherlands, UK, Spain and Germany) to 12% (Bulgaria and the Czech Republic).

image

Figure 1. Proportion (%) of reported carbohydrate intake by glycaemic index confidence level at baseline.

Download figure to PowerPoint

Figure 2 illustrates the proportion of carbohydrate intake by GI assignment level during the dietary intervention, when participants had been given specific advice to follow a low-fat diet and on specific carbohydrate-rich foods with known GI values to lower/raise dietary GI. Accordingly, the proportion of carbohydrate intake from confidence level 1 and 2 foods was higher than at baseline, but inclusion of confidence level 3 foods resulted in similar values to those obtained at baseline.

image

Figure 2. Proportion (%) of reported carbohydrate intake by glycaemic index confidence level during dietary intervention (diaries from all post-baseline timepoints combined).

Download figure to PowerPoint

Table 4 shows the effect of changing the value assigned to the confidence level 5 foods. The selected nominal value of 70 was replaced by the mean dietary GI as calculated when excluding all confidence level 5 foods (centre-specific, GImean) and also by zero (GI0). As the mean dietary GI was lower than 70 in all centres, replacing 70 with this value reduced the resulting mean dietary GI calculated. Replacing the nominal value with 0 reduced it still further. The effect of the nominal value selected on dietary GI was greatest in Bulgaria and the Czech Republic, consistent with the larger proportion of carbohydrates assigned confidence level 5 in these centres.

Table 4.  The effect on dietary GI of assigning different values to confidence level 5 foods (analyses based on baseline data and presented by centre)
CentreMean GI*Mean dietary GI
GI70GImeanGI0
  • *

    Calculated excluding foods assigned GI level 5.

  • GI70 level 5 nominal value = 70.

  • GImean level 5 nominal value = mean centre-specific GI (see *).

  • GI0 level 5 nominal value = 0.

  • GI, glycaemic index.

Bulgaria52.761.459.253.2
Czech Republic53.160.558.452.3
Denmark60.364.363.861.0
Germany54.859.458.455.6
Greece57.162.260.957.2
Netherlands57.761.060.257.1
Spain54.357.756.853.9
UK56.660.559.756.8

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of Interest Statement
  8. Acknowledgement
  9. References

The methodology described here has enabled GI values to be assigned to around 6000 foods across five national food databases in a consistent manner. This has facilitated the determination of dietary GI and GL from food diaries alongside nutrient intakes. By including confidence levels together with assigned values in the databases, the methodology allows for assessment of the ‘quality’ of estimated dietary GI and GL. Generally, the greater the proportion of reported carbohydrate intake from foods assigned GI values with higher levels of confidence, the higher the quality of the estimated dietary GI and GL, whereas the higher the proportion of confidence level 4 and 5, the greater the uncertainty. This work has shown variation in the quality of the GI databases between European centres. This arises largely from the fact that the bulk of GI measurement work has been carried out in North America and Australia, and while a number of European foods have been tested, many foods typical of a number of the European countries have not been measured previously. Including assignment levels in the database also means that it can readily be seen where gaps lie in the available data, and hence where confidence can be improved in a value by the measurement of additional important foods. This allows for the database to be updated with higher confidence values as further GI values are measured and published.

It is the intention that this work will provide a foundation for the development of a European nutrient database incorporating GI values. Making the database available to other researchers in future will save others repeating a time-consuming process and provide consistency and comparability between studies. It will also help to identify priorities for future measurements and avoiding unnecessary repetition of measurement of foods. When deciding the priorities it is not enough to identify foods with low confidence levels, it is also necessary to take into account the contribution of foods to the carbohydrate intake. Examination of lists of foods assigned confidence levels 4 and 5 values together with respective intake data for these foods will reveal those that contribute most to carbohydrate intakes, and so have the greatest impact on dietary GI. For example, Fig. 2 reveals that in the Czech Republic, a relatively high proportion (12.5%) of reported carbohydrate intake was from confidence level 5 foods. Table 2 shows that just 115 foods in the Czech database were assigned confidence level 5 values. In combination, these data inform decision-making on which foods could most usefully be measured to improve the estimate of dietary GI. However, caution must be taken in assuming that a food that does not contribute significantly to carbohydrate intake on a population level is not important when assessing diets of individuals, as an individual may consume large quantities of that food.

There are a number of limitations to this work. In part, these relate to the generic issues of assigning GI values to reported foods. While the rules were determined in order to provide consistency and objectivity to the GI assignment, it is inevitable that some subjectivity is involved in the decisions. There remains the potential for foods to be assigned values that are incorrect, including at confidence level 2. This will decrease as more foods have measured GI values and can be assigned values at confidence level 1. In some cases, either the database food descriptions or the published GI values lacked detail regarding the form, preparation and/or cooking methods of the food. These factors are all known to influence the GI of a food (24–28). Certain assumptions had to be made in these situations based on available information. However, where it was felt that such an assumption could not be made with sufficient confidence, the confidence level 5 value was assigned. A number of studies have demonstrated that the GI of a mixed meal can be predicted from the GI of the constituent foods (22,29,30). However, this finding is not universal, and there is some doubt when protein and fat differ in addition to the carbohydrate component (31). In this work therefore, this assumption of predictability was only made when the sole difference between foods or meals was in the carbohydrate component. Some previous studies have attempted to calculate a ‘precise’ predicted GI based on proportions of ingredients, rather than using the closest likely estimate approach taken (5,7). We believe this to be a more pragmatic approach given the multiple sources of uncertainty and variability involved in predicting values.

Another source of uncertainty arises from assumptions made in the coding of diaries where insufficient information is given by respondents, and which may have substantial impact on the dietary GI. For example, if rice is recorded without the type being specified, the diet coder must select a food from the list available to them in the database. As different types of rice have broadly similar nutrient compositions, this subjective selection will have little impact on the nutrient output. However, different types of rice have very different GI values, meaning that selection of the wrong type could cause a substantial error in the GI. Other aspects of foods, which are generally unimportant for nutrient composition but which can influence the GI, such as ripeness of fruit, are also not readily differentiated between in typical food diaries or nutrient databases. In order to minimize these sources of error, participants in interventions focused on GI should be instructed in the importance of being specific about these factors in their food diaries, and diet coders should be educated on the influence of these food factors on GI.

The selection of the nominal value requires further consideration. The dietary GI calculated excluding confidence level 5 foods ranged from 53 to 60 across countries. This means that the selection of 70 as the nominal value raises the dietary GI. A nominal value closer to these mean values, such as 60, may therefore be more appropriate to avoid biasing dietary GI estimates. Using the mean value for each country is initially attractive, but not a practical approach, as this will inevitably vary between studies and cannot be decided a priori. Variations in the nominal value could be accommodated in an online database by allowing for study-specific nominal values based on the typical dietary GI of the country or population being investigated, to be applied to all foods marked as confidence level 5.

Methodologies previously described for GI assignment have usually supported the analysis of dietary questionnaires such as food frequency questionnaires. In such dietary assessment methods, foods are grouped and GI values must therefore be allocated to represent an average of all foods included in the groupings. In this project, using food diaries, individual foods are treated separately and must all have values assigned to them. European researchers have previously assigned GI values to foods for analyses of food diaries or 24-h recalls. For example, the European Investigation into Cancer and Nutrition study assigned GI values to a European food database for calculation of dietary intakes from 24-h recalls (6). This work, in which GI and GL calculations from country specific dietary questionnaires were compared with single 24-h recalls, uses a slightly different approach from that described here. However, this and other study-specific databases have not been made publicly available. Future work should be addressed at comparing dietary GI and GL calculations using these two databases to determine where differences may arise and agreeing a common procedure. Understanding the factors contributing to different GI/GL results would help us to improve this GI database for future studies and may help to reduce one potential source of variability when comparing the health consequences of diets of higher or lower GI.

Acknowledgement

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of Interest Statement
  8. Acknowledgement
  9. References

Research relating to this article was funded by the EU (contract #: Food-2005-CT-513946; Diogenes number 1.54).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of Interest Statement
  8. Acknowledgement
  9. References
  • 1
    Van Dam RM, Seidell JC. Carbohydrate intake and obesity. Eur J Clin Nutr 2007; 61(Suppl. 1): S75S99.
  • 2
    Jenkins DJ, Wolever TM, Taylor RH, Barker H, Fielden H, Baldwin JM, Bowling AC, Newman HC, Jenkins AL, Goff DV. Glycemic index of foods: a physiological basis for carbohydrate exchange. Am J Clin Nutr 1981; 34: 362366.
  • 3
    FAO/WHO. Carbohydrates in Human Nutrition. Report of a Joint FAO/WHO Expert Consultation. FAO Food and Nutrition Paper 66. FAO/WHO: Rome, 1998.
  • 4
    Willett W, Manson J, Liu S. Glycemic index, glycemic load, and risk of type 2 diabetes. Am J Clin Nutr 2002; 76: 274S280S.
  • 5
    Flood A, Subar AF, Hull SG, Zimmerman TP, Jenkins DJ, Schatzkin A. Methodology for adding glycemic load values to the National Cancer Institute Diet History Questionnaire database. J Am Diet Assoc 2006; 106: 393402.
  • 6
    Van Bakel MM, Slimani N, Feskens EJ, Du H, Beulens JW, Van Der Schouw YT, Brighenti F, Halkjaer J, Cust AE, Ferrari P, Brand-Miller J, Bueno-de-Mesquita HB, Peeters P, Ardanaz E, Dorronsoro M, Crowe FL, Bingham S, Rohrmann S, Boeing H, Johansson I, Manjer J, Tjonneland A, Overvad K, Lund E, Skeie G, Mattiello A, Salvini S, Clavel-Chapelon F, Kaaks R. Methodological challenges in the application of the glycemic index in epidemiological studies using data from the European Prospective Investigation into Cancer and Nutrition. J Nutr 2009; 139: 568575.
  • 7
    Simila ME, Valsta LM, Virtanen MJ, Hatonen KA, Virtamo J. Glycaemic index database for the epidemiological Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study. Br J Nutr 2008; 16.
  • 8
    Moore CS, Lindroos AK, Kreutzer M, Larsen TM, Astrup A, Van Baak MA, Handjieva-Darlenska T, Hlavaty P, Kafatos A, Kohl A, Martinez JA, Monsheimer S, Jebb SA. Dietary strategy to manipulate ad libitum macronutrient intake, and glycaemic index, across eight European countries in the Diogenes Study. Obes Rev 2009; June 1, DOI:10.1111/j.467-789X.2009.00602.x.
  • 9
    Larsen TM, Dalskov S, Van Baak M, Jebb S, Kafatos A, Pfeiffer A, Martinez JA, Handjieva-Darlenska T, Kunesova M, Holst C, Saris WH, Astrup A. The Diet, Obesity and Genes (Diogenes) Dietary Study in eight European countries – a comprehensive design for long-term intervention. Obes Rev 2009; May 21, DOI:10.1111/j.467-789X.2009.00603.x.
  • 10
    FSA. McCance & Widdowson's the Composition of Foods, 6th summary edn. Royal Society of Chemistry: Cambridge, 2002.
  • 11
    Institute DNF. The Danish Food Composition Databank, revision 5.0. [WWW document]. URL http://www.foodcomp.dk/fvdb_default.htm (accessed November 2009).
  • 12
    Voedingscentrum. NEVO-Tabel: Nederlands Voedingsstoffenbestand/Stichting Nederlands Voedingsstoffenbestand. Voedingscentrum: Den Haag, 2001.
  • 13
    Nutri-Science. PRODI® Version 4.5 (Based on the Bundeslebensmittelschlüssel). Nutri-Science: Hausach, 2001.
  • 14
    Alimenta. Databáze složení potravin version 4.0. Výskumnýústav potravinársky, 2001.
  • 15
    Wolever TM, Vorster HH, Bjorck I, Brand-Miller J, Brighenti F, Mann JI, Ramdath DD, Granfeldt Y, Holt S, Perry TL, Venter C, Xiaomei W. Determination of the glycaemic index of foods: interlaboratory study. Eur J Clin Nutr 2003; 57: 475482.
  • 16
    Aston LM, Gambell JM, Lee DM, Bryant SP, Jebb SA. Determination of the glycaemic index of various staple carbohydrate-rich foods in the UK diet. Eur J Clin Nutr 2007; 62: 279285.
  • 17
    Foster-Powell K, Holt SH, Brand-Miller JC. International table of glycemic index and glycemic load values: 2002. Am J Clin Nutr 2002; 76: 556.
  • 18
    Henry CJ, Lightowler HJ, Strik CM, Renton H, Hails S. Glycaemic index and glycaemic load values of commercially available products in the UK. Br J Nutr 2005; 94: 922930.
  • 19
    Henry CJ, Lightowler HJ, Strik CM, Storey M. Glycaemic index values for commercially available potatoes in Great Britain. Br J Nutr 2005; 94: 917921.
  • 20
    The University of Sydney. GI Database. [WWW document]. URL http://www.glycemicindex.com (accessed November 2009).
  • 21
    Wolever TM, Brand-Miller JC, Abernethy J, Astrup A, Atkinson F, Axelsen M, Bjorck I, Brighenti F, Brown R, Brynes A, Casiraghi MC, Cazaubiel M, Dahlqvist L, Delport E, Denyer GS, Erba D, Frost G, Granfeldt Y, Hampton S, Hart VA, Hatonen KA, Henry CJ, Hertzler S, Hull S, Jerling J, Johnston KL, Lightowler H, Mann N, Morgan L, Panlasigui LN, Pelkman C, Perry T, Pfeiffer AF, Pieters M, Ramdath DD, Ramsingh RT, Robert SD, Robinson C, Sarkkinen E, Scazzina F, Sison DC, Sloth B, Staniforth J, Tapola N, Valsta LM, Verkooijen I, Weickert MO, Weseler AR, Wilkie P, Zhang J. Measuring the glycemic index of foods: interlaboratory study. Am J Clin Nutr 2008; 87: 247S257S.
  • 22
    Chew I, Brand JC, Thorburn AW, Truswell AS. Application of glycemic index to mixed meals. Am J Clin Nutr 1988; 47: 5356.
  • 23
    Mayer-Davis EJ, Dhawan A, Liese AD, Teff K, Schulz M. Towards understanding of glycaemic index and glycaemic load in habitual diet: associations with measures of glycaemia in the Insulin Resistance Atherosclerosis Study. Br J Nutr 2006; 95: 397405.
  • 24
    Aston LM. Glycaemic index and metabolic disease risk. Proc Nutr Soc 2006; 65: 125134.
  • 25
    Jenkins DJ, Wesson V, Wolever TM, Jenkins AL, Kalmusky J, Guidici S, Csima A, Josse RG, Wong GS. Wholemeal versus wholegrain breads: proportion of whole or cracked grain and the glycaemic response. BMJ 1988; 297: 958960.
  • 26
    Kay RM, Stitt S. Food form, postprandial glycemia, and satiety. Am J Clin Nutr 1978; 31: 738739.
  • 27
    Bjorck I, Granfeldt Y, Liljeberg H, Tovar J, Asp NG. Food properties affecting the digestion and absorption of carbohydrates. Am J Clin Nutr 1994; 59: 699S705S.
  • 28
    Collier G, McLean A, O'Dea K. Effect of co-ingestion of fat on the metabolic responses to slowly and rapidly absorbed carbohydrates. Diabetologia 1984; 26: 5054.
  • 29
    Wolever TM, Bolognesi C. Prediction of glucose and insulin responses of normal subjects after consuming mixed meals varying in energy, protein, fat, carbohydrate and glycemic index. J Nutr 1996; 126: 28072812.
  • 30
    Wolever TM, Jenkins DJ. The use of the glycemic index in predicting the blood glucose response to mixed meals. Am J Clin Nutr 1986; 43: 167172.
  • 31
    Flint A, Moller BK, Raben A, Pedersen D, Tetens I, Holst JJ, Astrup A. The use of glycaemic index tables to predict glycaemic index of composite breakfast meals. Br J Nutr 2004; 91: 979989.