The group evaluated the Dietary Guidelines from the United States Department of Agriculture (USDA) (3–5) and the Dietary Reference Values from EFSA (6), both published in 2010. Furthermore, an updated report on Carbohydrates in Human Nutrition from WHO published 2007 (7), a report from The World Cancer Research Fund (8), and a review on the GI from the Nordic Council of Ministers from 2005 (9) was included in the review. When reporting conclusions from these reports, same or similar wording is used.

For the original articles from the Nordic countries, we defined the literature search and criteria for inclusion and exclusion, set prior to abstract screening. The eligibility criteria is similar to the criteria used in the systematic review on sugar and cardiometabolic risk factors by the same authors (2) and were based on the following aspects:

Postprandial glycemia refers to the elevation of blood glucose concentrations after consumption of a food or a meal, and is a normal physiological response which varies in magnitude and duration. The concept of GI lists food by its proposed effect on postprandial blood glucose compared to a reference food, or more recently, to pure glucose. GI can be influenced by the chemical and physical nature of the food or meal consumed but also by individual factors (27). These factors include nutrients such as certain types of DF, starch structure, type of sugar, fat and protein content, water, cellular structure and other structure related factors, molecular interactions, particle size distribution, presence of amylase inhibitors or organic acids, method of food preparation and degree of chewing (9).

The concept of the GI was developed in 1981, mainly in relation to people with diabetes (28). Current guidelines for the treatment of people with diabetes include the GI as a helpful additional indicator regarding the appropriate carbohydrate containing foods for inclusion in the diet (29, 30). Studies have also examined whether choosing food according to its GI could prevent the development of different diseases or metabolic conditions. The results are not consistent and methodological challenges have been pointed out as a potential explanation.

First, considerable individual differences in glucose responses can be found when testing various food items. Current guidelines on the number of persons necessary for testing a food item have been debated (27) as well as age and training state of subjects (31, 32). Second, GI is measured using 50 g of available carbohydrates from both the test food and the reference food (glucose). However, the availability of carbohydrates in some food items (for example due to resistant starch) may be overestimated resulting in lower amount of carbohydrates absorbed that might partly explain lower GI (33). This should however be over-come by analyzing the available carbohydrate portion. The glycemic index should only be used to rank food items with at least 10–20 g of available carbohydrates in a portion of the food (9).

Third, many cohort studies do not measure the GI of food in their country, but instead rely on international table values, which may not be suitable for local food items (34) and the comprehensiveness of the dietary questionnaires in epidemiological studies varies (27). Fourth, controlling for the effects of different types of food, DF and other macronutrients is difficult in both epidemiological studies and randomized controlled trials, and in many trials, low-GI and control diets generally differ in other potentially confounding aspects, such as soluble and insoluble fiber content, or viscosity of fiber and other macronutrient content. It can be hard to disentangle the positive effect due to lower glycemic and insulinemic response from positive effect due to other qualities of the food (27, 33). Studies on fiber have similar challenges.

Fifth, the GI does not take onto account the ‘second meal effect’ (the effect of a previous meal on the postprandial glycemia of a second meal) and meal frequency, which are important factors influencing the glycemic response (1). These can be controlled for in intervention studies but can rarely be accounted for in cohort studies assessing the association of the GI with health outcomes. Sixth, the extent to which GI for individual food items predicts glucose responses as part of a mixed meal has been discussed (27), but very frequently it can be predicted from the component carbohydrate moieties. The GI might be most accurate when comparing a low GI food item with a high GI food item, without making any other changes in composition of the meal (35, 36). Last, because there is not always coherence between the GI and insulinemic response, some researchers have used the insulin index as well (36–38). However, the methodology of GI analysis and study design has been steadily improving and issues resolved (27, 33, 34, 39).

The association between the GI and body weight has been covered in a systematic review on macronutrients, food consumption and weight changes, within the NNR project (40).

In the Report of the Dietary Guidelines Advisory Committee on the Dietary Guidelines for Americans, 2010 (13) the chapter on carbohydrates is partly based on systematic reviews. For GI and GL, four scientific questions were asked by the Committee (Table 4).

The report concludes, ‘When selecting carbohydrate foods, there is no need for concern with their glycemic index or glycemic load. What is important to being mindful of their calories, caloric density, and fiber content.’

The opinion by EFSA gives a review (not a systematic literature review) of studies investigating the GI or GL and glucose tolerance and insulin sensitivity (13 studies), serum lipids (four original studies, and one meta-analysis based on 15 studies), body weight (11 interventions, four cohorts studies and one review), type 2 diabetes (11 studies, one meta-analysis), cardiovascular disease (six studies and one systematic review) and colorectal cancer (meta-analysis based on 11 studies). It concludes that ‘Although there is some support for a role of GI and GL in the treatment of type 2 diabetes and some evidence suggesting that lowering GI and GL may have favorable effects on some metabolic risk factors such as serum lipids, the evidence regarding their role in the prevention of diet-related diseases is still inconclusive.’6

In 2011, the same EFSA panel published a scientific opinion on the substantiation of health claims related to reduction of post-prandial glycemic responses for different sugar replacers (xylitol, sorbitol, mannitol, maltitol, lactitol, isomalt, erythritol, D-tagatose, isomaltulose, sucralose, or polydextrose). The opinion concluded that reduction of post-prandial glycemic responses (as long as post-prandial insulinemic responses are not disproportionally increased) may be a beneficial physiological effect, for example to subjects with impaired glucose tolerance, common in the general population of adults.

The previous report from FAO/WHO Expert Consultation published in 1998 suggested that the concept of GI might provide useful means of helping to select the most appropriate carbohydrate containing foods for the maintenance of health and the treatment of several diseases (41). The current update from 2007 does not take as this as clear and points out some of the weaknesses of the GI concept and strongly expresses that a choice of carbohydrate rich foods could not be done on the basis of GI alone. GI is perhaps most appropriately used to guide food choices when considering similar carbohydrate-containing foods, for example bread with a low GI may be preferable to a higher GI bread, with a resultant lower GL, but should always be considered in the context of other nutritional factors. Furthermore, the report points out that although many manufactured food items have been tested for GI, most of the studies which have demonstrated a health benefit of low GI involved the use of naturally occurring and minimally processed foods. They suggest that manufactured foods be further examined, rather than to assume health benefits only on the basis of their functionality, i.e. their glycemic response. The update concludes, however, that despite the reservations made, distinguishing between foods with large enough differences in GI may produce some benefit in terms of glycemic control in diabetes and lipid management. Further, that low dietary GL may reduce the risk of type 2 diabetes and cardiovascular disease, but the same caveats as described above should be used when interpreting such observations.

The systematic review and meta-analysis from the World Cancer Research Fund published in 2010 is an update based on the previous Expert Report published 2007 (42). The updated report found limited evidence for an association between the GI of the diet and cancer risk (breast and colorectal cancer), and no conclusion was drawn (43, 44). However, the report did not elaborate on the total number, type or quality of studies reviewed for their findings.

The Tema Nord report published in 2005 focus on the GI and GL and their associations with metabolic risk factors and diseases as well as overweight/obesity and satiety (9). The report is not a systematic review but includes more than 100 papers. The report concluded that:

For individuals with diabetes or impaired glucose tolerance a low-GI diet might be of importance; this holds as well for those prone to diabetes due to overweight. More evidence is needed to be able to draw more secure conclusions on the importance of low GI food for healthy individuals.

From the 35 identified abstracts, 12 relevant studies were included. The included studies are summarized in Table 5. One study examined the relation with serum lipids, one study examined cardiovascular disease, one study examined heart failure, three studies examined myocardial infarction (MI) and one study examined type 2 diabetes. The remaining five studies examined the relation with different types of cancers, breast cancer (two studies), stomach cancer (one study), endometrial cancer (one study) and colorectal cancer (one study). All of the papers were prospective cohort studies.

No association was found between the estimated GI or GL of the diet and risk for type 2 diabetes (45), ischemic cardiovascular disease (46) or heart failure (55). Out of the three studies on MI, one study did not find an association (47), one study reported a positive association only among obese or physically inactive individuals (56), and the third study reported a lower risk of MI if replacing saturated fatty acids with low GI food instead of high GI food on equal percentage of energy basis (48). The study on serum lipids found an association albeit weak and generally in sub-groups as described before (49). No association was found with stomach cancer (50) or colorectal cancer (51), while a positive association was found between GI and endometrial cancer only among overweight women with low physical activity (52). For breast cancer, one of the two studies (53, 54) found an overall weak positive association with GL (53) while both studies found positive associations with sub-types of breast cancer.

Additionally, four Nordic studies on the GI concept, which did not fit the eligibility criteria, are of interest. Two of the studies focused on ways to change the GI of different food items using resistant starch (chilled potatoes) and vinegar (57, 58), a study finding consumption of barley kernels compared to white wheat bread in the previous evening meal to lower the glucose tolerance when giving the same breakfast to both groups of healthy volunteers (59), and one cross-over study finding lower postprandial glucose and insulin response in healthy subjects when adding 4 g of oat β-glucans to a bread compared with no addition (60). All efforts were made to identify relevant Nordic studies. However, not all researchers mark their studies (index, title or abstract) by country and therefore relevant studies from the Nordic countries might have been missed.