Glycaemic index methodology

A very high correlation exists between the rate of in vitro glucose release from starchy foods using pancreatic and brush-border enzymes and the glycaemic response in vivo.

“A very high correlation exists between the rate of in vitro glucose release from starchy foods, using pancreatic and brush-border enzymes, and the glycaemic response in vivo (Granfeldt et al. 2005;Englyst et al. 2003).”

• cites Granfeldt et al. 2005Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Englyst et al. 2003 sourcePartly supported

  The Dietary carbohydrate analysis section measures glucose released during in vitro enzymatic hydrolysis, and the Discussion reports strong correlations between those measures and glycaemic index. However, the study tested 23 cereal products, not starchy foods generally, and our copy does not specify that the enzyme mixture contained both pancreatic and brush-border enzymes. (LLM, conf=0.86)

  Experimental — accuracy not yet measured

According to Brand-Miller (2004), carbohydrate quality as glycaemic index more often shows a significant association with disease risk, including diabetes, cardiovascular disease and cancer, than does carbohydrate content or glycaemic load.

“According to Brand-Miller (2004), the quality of CHO (i.e.GI) more often shows a significant association with disease risk (diabetes, CVD, cancer) than does the CHO content or GL of the diet.”

• cites Brand-Miller (2004) sourceSupported

Accumulating evidence indicates that diets containing a preponderance of low-GI foods induce modest to clinically important benefits in the intermediate term as shown by intervention studies.

“there is accumulating evidence (Brand-Miller, 2004) that diets containing a preponderance of foods that elicit low glycaemic responses ('low-GI foods or diets'), as originally defined by Jenkins et al. (1981) induce modest to clinically important benefits in the intermediate term as shown by intervention studies”

• cites Brand-Miller, 2004 sourcePartly supported

  In the main text (page unavailable), Brand-Miller says ample clinical and other evidence supports minimizing postprandial glycemia. But it does not report that intervention studies of diets predominantly containing low-GI foods produced modest to clinically important benefits over an intermediate term; it describes the use of low-GI carbohydrates as controversial and relies chiefly on observational findings and an acarbose trial. (LLM, conf=0.83)

  Experimental — accuracy not yet measured
• cites Jenkins et al. (1981) sourceNot checked — no full text of the source

The average of three trials of the reference food reduces the variation of mean glycaemic index values.

“The average of three trials of the reference food has been shown to reduce the variation of mean GI values (Wolever et al. 1991(Wolever et al. , 2002)).”

• cites Wolever et al. 1991 sourceNot checked — no full text of the source
• cites Wolever et al. 2002 sourceCould not verify

  The Methods describe using food records to give dietary advice, and the Results report dietary GI and glucose disposition outcomes. Neither section addresses trials of a reference food or whether averaging three trials reduces variation in mean GI values (searched: reference food, three trials, and variability). (LLM, conf=0.76)

  Experimental — accuracy not yet measured

The twenty-three cereal-based foods had very similar contents of non-available carbohydrate.

“According to the results of analytical measurement of NSP and RS contents the twenty-three foods had very similar content in non-available CHO.”

• cites Englyst et al. (2003) sourcePartly supported

  The Test meals section identifies twenty-three cereal products, and the Dietary carbohydrate analysis section describes measurement of resistant starch. But the Relationships between nutrients section does not establish that their combined non-available carbohydrate contents were very similar; it notes an exceptionally high NSP value for Alpen and a range of RS values. Page numbers are unavailable in our copy. (LLM, conf=0.78)

  Experimental — accuracy not yet measured

In several comparative studies, the glycaemic impact of carbohydrate foods has been related to the rate of digestion of available carbohydrate.

“In several comparative studies, the glycaemic impact of CHO foods has been related to the rate of digestion of available CHO.”

• cites Englyst et al. (2003) sourceSupported

Other studies have observed or indicated a relationship between the volume of water or food and postprandial responses, including incremental areas.

“whereas others observed or indicate a relationship between the volume of water or food and postprandial responses, and sometimes incremental areas (Torsdottir & Anderson, 1989;Young & Wolever, 1998;Sievenpiper et al. 2001).”

• cites Torsdottir & Anderson, 1989 sourceSupported
• cites Young & Wolever, 1998 sourceNot checked — no full text of the source
• cites Sievenpiper et al. 2001 sourceNot checked — no full text of the source

Studies that have examined how volume affects glycaemic response have produced contradictory results.

“Unfortunately, the results of studies undertaken to date to evaluate the effect of volume on glycaemic response are contradictory.”

• cites Gregersen et al. 1990 sourceNot checked — no full text of the source
• cites Torsdottir & Anderson, 1989 sourceCould not verify

  The introduction says earlier abstracts reported increased glycaemic responses when water was added to meals, and the Discussion reports an increase in two groups in this study but no significant effect in poorly controlled Type 2 diabetic patients. Those are differences between groups, not evidence that studies of meal volume produced contradictory results. No passage found makes that broader comparison across studies; page numbers are unavailable in our copy. (LLM, conf=0.76)

  Experimental — accuracy not yet measured
• cites Young & Wolever, 1998 sourceNot checked — no full text of the source

Englyst et al. correlated twenty-three cereal-based starchy foods for their in vivo GI and their in vitro rate of carbohydrate digestion, assessed as rapidly available glucose.

“Englyst et al. (2003) correlated twenty-three cereal-based starchy foods for their GI in vivo and their rate of CHO digestion in vitro (as assessed by the content in rapidly available glucose).”

• cites Englyst et al. (2003) sourceSupported

Most studies in this area compare the effects of exchanging sucrose or dextrins with fructo-oligosaccharides or inulin.

“Regarding the latter, most of the studies in this area compare the effects of exchanging sucrose or dextrins with fructo-oligosaccharides or inulin (Williams & Jackson, 2002).”

• cites Williams & Jackson, 2002 sourcePartly supported

  The section on lipid responses in human volunteers describes studies comparing oligofructose with sucrose and inulin with placebo. It does not identify dextrins as a comparator or establish that most studies exchanged sucrose or dextrins for these ingredients. (LLM, conf=0.83)

  Experimental — accuracy not yet measured

The effect of evening meal macronutrient composition on fat tolerance is much more marked than its effect on glucose tolerance.

“However, the effect of evening meal macronutrient composition on fat tolerance is much more marked than the effect on glucose tolerance (Robertson et al. 2002).”

• cites Robertson et al. 2002 sourcePartly supported

  The Oral-fat-tolerance test section reports differences in fatty acids and triacylglycerol after the two evening meals, while the Glucose and insulin concentrations section reports a significant difference in glucose response (P = 0.006). The paper does not establish that the effect on fat tolerance was much more marked than the effect on glucose tolerance; its Figure 3 caption describes the evening-meal effect on triacylglycerol during the fat-tolerance test as only nearly significant (P = 0.078). Page numbers are unavailable in our copy. (LLM, conf=0.86)

  Experimental — accuracy not yet measured

Intake of dietary resistant starch on the day before a meal challenge may improve glucose tolerance.

“Intake of dietary RS on the day before a meal challenge may improve glucose tolerance (Robertson et al. 2002).”

• cites Robertson et al. 2002 sourcePartly supported

  The Oral-glucose-tolerance test section reports improved next-morning glucose tolerance after a high-carbohydrate evening meal. But the Test meal composition and protocol section says the meals were very low in dietary fiber, and the paper does not identify or test resistant starch. The finding therefore supports the timing and glucose-tolerance outcome, not the claim about resistant starch specifically. Page numbers are unavailable in our copy. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

No differences have been observed in glycaemic response between males and females.

“No differences have been observed in glycaemic response between males and females (Wolever et al. 2002).”

• cites Wolever et al. 2002 sourceCould not verify

  Table 1 lists the men-to-women ratios of the study groups, but does not compare glycaemic responses between men and women. The Results report comparisons between glucose-tolerance and diet groups, not between sexes. Searches for sex-specific glucose responses found no supporting passage in our copy; page numbers were not available. (LLM, conf=0.73)

  Experimental — accuracy not yet measured

For many CHO-rich foods and pure CHO, there appears to be a close correlation between glycaemic and insulinaemic responses.

“However, for many CHO-rich foods and pure CHO, there appears to be a close correlation between glycaemic and insulinaemic responses (Holt et al. 1997;Englyst et al. 2003).”

• cites Holt et al. 1997 sourceNot checked — no full text of the source
• cites Englyst et al. 2003 sourcePartly supported

  The Discussion says the study confirms a relationship between glycaemic and insulinaemic indices for starchy foods, but describes the association as rather weak. It does not establish a close correlation across many carbohydrate-rich foods or for pure carbohydrates. Page numbers were not available in our copy. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

An insulin index is likely to be influenced by the pathophysiological status of the test subject.

“Moreover, an insulin index is likely to be influenced by the pathophysiological status of the test subject (Wolever et al. 2004).”

• cites Wolever et al. 2004 sourceSupported

When blood glucose falls, fingertip capillary glucose is lower than forearm.

“conversely, when blood glucose concentrations fall, fingertip capillary glucose concentrations are lower than those on the forearm (Jungheim & Koschinsky, 2002).”

• cites Jungheim & Koschinsky, 2002 sourceSupported

In an interlaboratory GI study, laboratories using venous blood returned CI of 50%, compared with 30% in laboratories using capillary blood.

“This has been confirmed directly in the interlaboratory study of GI, in which laboratories taking venous blood returned CI of .50 %, compared with , 30 % in all laboratories using capillary blood (Wolever et al. 2002(Wolever et al. , 2004)).”

• cites Wolever et al. 2002 sourceCould not verify

  The Methods and Results passages (page numbers unavailable) describe a dietary intervention and glucose disposition measurements, not an interlaboratory GI study. Searches for venous and capillary sampling, interlaboratory variation, and the 50% versus 30% confidence intervals found no passage addressing the claim. (LLM, conf=0.76)

  Experimental — accuracy not yet measured
• cites Wolever et al. 2004 sourceCould not verify

  The Research Design and Methods passages (page not supplied) describe a single study comparing responses to high- and low-fiber cereals, not an interlaboratory GI study. Searches for venous versus capillary blood and the claimed 50% versus 30% confidence intervals found no passage addressing that comparison in our copy. (LLM, conf=0.75)

  Experimental — accuracy not yet measured

In Flint et al. 2004, the GI of mixed breakfast meals calculated using GI table values did not correlate with GI measured in vivo.

“In their study they could not show that the GI of mixed breakfast meals, as calculated by using GI table values, correlated with the GI as measured in vivo.”

• cites Flint et al. 2004 sourceSupported

In Flint et al. 2004, a stronger relationship was observed for fat, protein, or energy content than for carbohydrate alone.

“A stronger relationship was observed for either fat or protein or energy content than for CHO alone.”

• cites Flint et al. 2004 sourcePartly supported

  The Abstract and 'Prediction of glycaemic index based on meal components' say that fat and protein together, or energy content, predicted glycaemic index more strongly than carbohydrate content alone. The claim changes 'fat and protein' to 'fat or protein': the 'Relationships between measured glycaemic index and meal variables' section does not support protein individually being stronger, reporting a lower R² for protein amount (0.65) than for carbohydrate as a percentage of energy (0.80). (LLM, conf=0.88)

  Experimental — accuracy not yet measured

Prolonging the time of ingestion from a few minutes to several hours markedly flattens the glycaemic response in normal and diabetic subjects.

“It is known for both solid and liquid test meals, in normal and diabetic subjects, that prolonging the time of ingestion from a few minutes to several hours markedly flattens the glycaemic response (Jenkins et al. 1990(Jenkins et al. , 1992)).”

• cites Jenkins et al. 1990 sourceNot checked — no full text of the source
• cites Jenkins et al. 1992 sourceNot checked — no full text of the source

The amount of drink taken by subjects differs from one study or laboratory to another, being a second source of variation of the volume.

“However, even if the amount of drink taken by the subjects is controlled, it differs from one study, or laboratory, to another, being a second source of variation of the volume.”

White bread from a certain local area may differ in glycaemic index compared with a common white bread from another area, as happened for the French white bread baguette.

“In contrast, white bread from a certain local area may differ in GI compared with a common white bread from another area, as happened for the French white bread baguette (Foster-Powell et al. 2002).”

• cites Foster-Powell et al. 2002 sourceNot checked — no full text of the source

The margin of error of the estimate of mean glycaemic index decreased substantially from one to two reference measurements.

“The margin of error of the estimate of mean GI decreased substantially from one to two reference measurements.”

Glycaemic index values obtained when white bread is used are typically about 1.4 times those obtained if glucose is used as standard.

“The GI values obtained when white bread is used are typically about 1•4 times those obtained if glucose is used as standard (Food and Agriculture Organization, 1998).”

• cites Food and Agriculture Organization, 1998Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

In the interlaboratory study, correlations between test food and reference food area under the curve measurements were typically 0.6 to 0.7.

“For the interlaboratory study (Wolever et al. 2003), correlations between test food and reference food AUC measurements were typically 0•6 to 0•7.”

• cites Wolever et al. 2003Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The standard deviation of glycaemic index values is linearly related to their means.

“The SD of GI values were linearly related to their means (Fig. 1).”

• cites Wolever et al. 2003Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

There is virtually no benefit from taking more than three to four measurements of the reference food area under the curve.

“There was virtually no benefit to be gained from taking more than three to four measurements of the reference food AUC.”

Brouns et al. (1995) observed a constant delivery of about 1.2 g carbohydrate/min for various solutions containing 45-90 g carbohydrate/l and osmolarity 243-374 mOsm/kg.

“For example Brouns et al. (1995) observed a constant delivery of about 1•2 g CHO/min for various solutions containing a range of 45 -90 g CHO/l and an osmolarity of 243-374 mOsm/kg.”

• cites Brouns et al. (1995) sourceNot checked — no full text of the source

Malabsorption accounted for only about 12% of the reduction in glycaemic response.

“Thus, malabsorption only accounted for about 12 % of the reduction in glycaemic response (100 £ 9/75 ¼ 12 %).”

Even at the high levels of carbohydrate malabsorption seen with legumes, malabsorption accounts for less than 20% of the reduced glycaemic response.

“Even at high levels of CHO malabsorption seen with legumes, the reduction in CHO malabsorption accounts for , 20 % of the reduced glycaemic response.”

The difference in absorbed carbohydrate between bread and lentils would account for about a 9% difference in glycaemic response.

“This difference in absorbed CHO would account for about a 9 % difference in glycaemic response (100 £ (45 2 41)/45 ¼ 8•9 %).”

Incremental AUC accounted for only 14 to 38% of total AUC.

“For example, from Table 2, it can be seen that incremental AUC accounted for only 14 to 38 % of total AUC.”

Analysing more than one baseline sample had only a very small effect overall on incremental AUC in data from nineteen subjects.

“In practice, analysing data from nineteen subjects showed this to have only a very small effect overall on the incremental AUC.”

Mean GI values measured in different laboratories vary.

“Mean GI values measured in different laboratories vary.”

AUC following white bread was related to a substantially higher insulin response.

“AUC following white bread was related to a substantially higher insulin response.”

The highest precision of glycaemic index values is expected when within-subject variation of the incremental area under the curve of the blood glucose response is minimized.

“Theoretically, the highest precision of GI values is expected when the within-subject variation of the incremental AUC of the blood glucose response is minimised.”

A clinically significant decrease in protein glycation was observed with a reduction in dietary glycaemic load.

“Brand-Miller et al. (2003a) observed a clinically significant decrease of protein glycation with a reduction of GL of the diet.”

• cites Brand-Miller et al. (2003a)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

There is a linear relationship between the carbohydrate content of a drink and gastric emptying time.

“It was shown that there is a linear relationship between CHO content of the drink and gastric emptying time.”

The carbohydrate concentration and osmolarity of test drinks, at least in the range tested, are not expected to have any impact on glycaemic response.

“Accordingly, the CHO concentration and osmolarity of the test drinks, at least in the range tested, are not expected to have any impact on glycaemic response.”

The classical method of measuring dietary fibre did not measure resistant starch, non-digestible oligosaccharides and sugar alcohols appropriately, leading to underestimation of the true content of unavailable carbohydrate in foods and diets.

“Yet the classical method of measuring dietary fibre did not measure these CHO appropriately, leading to an underestimation of the true content of unavailable CHO in foods and diets.”

• cites Champ et al. 2003Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The rate of delivery of carbohydrate from the stomach to the small intestine is fairly constant for soluble available carbohydrate.

“Because of this, the rate of delivery of CHO from the stomach to the small intestine is fairly constant for soluble available CHO.”

GI was better explained by the rate of digestion of the available carbohydrate fraction than by the content of non-available carbohydrate, such as dietary fibre and resistant starch.

“Accordingly, GI was more explained by the rate of digestion of the available CHO fraction than by the content in non-available CHO (dietary fibre and RS).”

The blood glucose response pattern after lentils was followed much more closely by the continuous meal than by the quarter bread meal.

“However, the pattern of blood glucose response after lentils (Fig. 8) was followed much more closely by the continuous meal than by the quarter bread meal (Fig. 9).”

• cites Jenkins et al. (1982) sourceNot checked — no full text of the source

For high-amylose maize bread, the glycaemic response to a 45 g total starch portion was identical to that after a 55% larger portion accounting for 35% resistant starch.

“In the case of high-amylose maize bread, the glycaemic response elicited by the 45 g total starch portion, 49 mmol £ min/l, was identical to the response after a 55 % larger portion size to account for the 35 % RS, 49 mmol £ min/l (Granfeldt et al. 1995a).”

• cites Granfeldt et al. 1995aNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Caffeine is known to acutely decrease insulin sensitivity in human subjects.

“Moreover, caffeine is known to acutely decrease insulin sensitivity in human subjects (Graham et al. 2000;Keijzers et al. 2002).”

• cites Graham et al. 2000Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Keijzers et al. 2002 sourceNot checked — no full text of the source

For barley, an 18% increase in portion size produced a non-significant 13% increase in GI from 39 to 44.

“In the case of barley, an 18 % increase in portion size resulted in a non-significant 13 % increase in GI from 39 to 44 (glucose ¼ 100; Wolever et al. 2003).”

• cites Wolever et al. 2003Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Glycaemic response also depends on the amounts and types of fat, protein, fibre and other compounds and their interactions.

“Thus it also depends on the amounts and types of fat, protein, fibre and other compounds and their interactions.”

The major determinant of the low GI values of starchy foods in this study is a reduced rate of carbohydrate absorption.

“It is therefore believed that the major determinant of the low GI values of starchy foods in this study is a reduced rate of CHO absorption.”

Similar conclusions were drawn from second-meal experiments by Liljeberg et al.

“Similar conclusions were drawn from second-meal experiments done by Liljeberg et al. (1999).”

• cites Liljeberg et al. (1999) sourceNot checked — no full text of the source

The truly available carbohydrate fraction is difficult to determine with in vitro models because they do not account for gastrointestinal influences, such as in vitro digestible carbohydrate escaping small-intestinal digestion and ending unabsorbed in the colon.

“The truly available CHO fraction is difficult to determine by in vitro models, since these do not account for any influences within the gastrointestinal tract, for example the fact that some in vitro digestible CHO may escape small-intestinal digestion and absorption, thus ending unabsorbed in the colon.”

The glycaemic response elicited by lentils was 75% less than that of bread.

“In fact, however, the glycaemic response elicited by lentils was 75 % less than that of bread (Wolever et al. 1986a).”

• cites Wolever et al. 1986aNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

In 2002 AOAC recognised a new method that measures type 3-retrograded resistant starch specifically and accurately.

“However, in 2002 AOAC recognised an additional new method (McCleary & Monaghan, 2002) that measures type 3-retrograded RS specifically and accurately.”

• cites McCleary & Monaghan, 2002 sourceNot checked — no full text of the source

Short-term metabolic effects of reducing available carbohydrate intake differ from those of slowing the rate of carbohydrate absorption.

“However, it has been shown that short-term metabolic effects of reducing available CHO intake are different from those of slowing the rate of CHO absorption.”

About 10% of the available carbohydrate in bread and about 18% in lentils was malabsorbed, according to both breath hydrogen and ileal effluent measurements.

“Both methods indicated that about 10 % of the 'available' CHO in bread and about 18 % in lentils was malabsorbed.”

• cites Wolever et al. (1986a)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Several enzymic procedures simulate gastrointestinal digestion and allow analysis of available and resistant starch content in foods.

“There are also several enzymic procedures available simulating the digestion in the gastrointestinal tract and making it possible to analyse the available and RS content of foods (Englyst et al. 1996;1998).”

• cites Englyst et al. 1996;1998Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The initial volume of a fluid strongly affects the rate at which the stomach empties.

“Fluids have a strong effect of initial volume on the gastric emptying rate.”

• cites Brouns, 1998 sourceNot checked — no full text of the source

Slowing the rate of carbohydrate absorption, for example by sipping glucose slowly over 3 hours, reduces glucose and insulin responses.

“Of course other mechanisms are possible, but it is true that slowing the rate of absorption of CHO, for example by sipping glucose slowly over 3 h, reduces glucose and insulin responses (Jenkins et al. 1990).”

• cites Jenkins et al. 1990 sourceNot checked — no full text of the source

An unknown and significant fraction of native starch may become resistant to digestion after preparation for consumption.

“This analytical progress seems of relevance since an unknown and significant fraction of native starch, once it has been prepared for consumption, may be resistant to digestion.”

The resistant starch content of a starch-based food may range from a few percent to as much as 40%, depending on the native source and meal preparation.

“As a consequence, the amount of RS in a starch-based food may range from a few percent to as much as 40 % depending on the native source and/or the way of meal preparation (Champ et al. 2003).”

• cites Champ et al. 2003Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Measuring available carbohydrate is no longer a problem for most common foods.

“Therefore, we can conclude that measuring available CHO is no longer a problem for most common foods.”

Some studies have found no significant effect of water volume on glycaemic response.

“Some authors found no significant effect of water volume (Gregersen et al. 1990)”

• cites Gregersen et al. 1990 sourceNot checked — no full text of the source

Fluids have a strong effect of initial volume on the gastric emptying rate.

“Fluids have a strong effect of initial volume on the gastric emptying rate.”

The second-meal response after lentils was 38% lower than after the control bread breakfast.

“The second-meal response after lentils was 38 % less than that after the control bread breakfast (Fig. 7).”

• cites Jenkins et al. (1982) sourceNot checked — no full text of the source

Analysis of total starch minus resistant starch by approved methods allows estimation of the available starch fraction.

“Analysis of total starch less RS obtained by approved methods makes possible estimation of the available starch fraction.”

Tea and coffee influence the pattern of the blood glucose response.

“however, as volume, they influence the pattern of blood glucose response.”

• cites Young & Wolever (1998) sourceNot checked — no full text of the source

The reason for the barley result is probably that the food factors responsible for intrinsic resistant starch formation also reduce the in vivo availability of the bulk of starch.

“The reason for this is probably that the food factors responsible for the intrinsic RS formation also reduce the in vivo availability of the bulk of starch.”

Effects seen when exchanging sucrose or dextrins with fructo-oligosaccharides or inulin may be due partly to short-chain fatty acid production during colonic fermentation.

“Effects that are seen in these cases may, along with a reduction of the amount of glycaemic CHO ingested, be due to the production of short-chain fatty acids during colonic fermentation (Brighenti et al. 1999).”

• cites Brighenti et al. 1999 sourceNot checked — no full text of the source

The effects of many prior-day measures are not of sufficient significance to warrant stricter controls in GI testing.

“Whilst the effects of many of these measures have been shown under other laboratory conditions, it appears that they are not of sufficient significance to warrant stricter controls when applied to GI testing.”

In starchy foods, the bioavailable starch fraction and resistant starch can be analysed with specific analytical procedures.

“In the case of starchy foods, the bioavailable starch fraction as well as RS can be analysed with specific analytical procedures (McCleary & Monaghan, 2002; for a review, see Champ et al. 2003).”

• cites McCleary & Monaghan, 2002 sourceNot checked — no full text of the source
• cites Champ et al. 2003Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Increasing carbohydrate intake increases incremental AUC glucose response in a dose-response manner in healthy subjects.

“Increasing the intake quantity of CHO increases incremental AUC glucose response in a dose -response manner in healthy subjects (Wolever & Bolognesi, 1996a;Lee & Wolever, 1998).”

• cites Wolever & Bolognesi, 1996aNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Lee & Wolever, 1998 sourceNot checked — no full text of the source

Not controlling for the amount of available carbohydrate would prevent valid conclusions about whether reduced glycaemic responses are due to a reduced amount of available carbohydrate or to another effect.

“Not controlling for the amount of available CHO would mean that valid conclusions as to whether reduced glycaemic responses were due to a reduced amount of available CHO or to some other effect cannot be drawn.”

For practical purposes, the available enzymic methods can be expected to give a reasonably good estimate of the available carbohydrate load in foods containing only intrinsic carbohydrate.

“For practical purposes, these methods can be expected to give a reasonably good estimate of the available CHO load in foods with only intrinsic CHO.”

There is evidence that glycaemic response data obtained at lunchtime after a standard breakfast differ significantly from data obtained after an overnight fast.

“There is evidence that the glycaemic response data obtained from a test at lunchtime, after a standard breakfast, differ significantly from those obtained after an overnight fast.”

Examples of indigestible carbohydrates not measured by the classical dietary fibre method include fructans, inulin, fructo-oligosaccharides and type 3 resistant starch.

“Some examples of such CHO are the fructans inulin and fructo-oligosaccharides and type 3 RS.”

• cites Champ et al. 2003Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

A particular glycaemic load may be achieved either from low-GI foods or from high-GI foods at a lower available carbohydrate intake.

“Thus, a particular GL may be achieved either from low-GI foods, or from high-GI foods at a lower available CHO intake.”

In Campbell et al. 2003, smokers returned similar glycaemic response and variability under controlled and uncontrolled conditions.

“Participants who were smokers (three out of thirteen) reported 'craving' during the controlled tests, but nevertheless returned similar values for glycaemic response and variability under the two conditions.”

• cites Campbell et al. 2003 sourceNot checked — no full text of the source

The carbohydrate content of the diet for a few days before an oral glucose tolerance test markedly affects the result, with poorer glucose tolerance after a low-carbohydrate diet.

“It has long been known that the CHO content of the diet for the few days before an oral glucose tolerance test markedly affects the result, with poorer glucose tolerance being observed after a low-CHO diet (Hales & Randle, 1963).”

• cites Hales & Randle, 1963 sourceNot checked — no full text of the source

After exercise, postprandial plasma glucose levels differed significantly from rest.

“Post-exercise, however, postprandial plasma glucose levels differed significantly from rest.”

• cites Folch et al. (2003) sourceNot checked — no full text of the source

In Campbell et al. 2003, mean glycaemic response did not differ between controlled and uncontrolled tests.

“There was no effect on mean values of glycaemic response.”

• cites Campbell et al. 2003 sourceNot checked — no full text of the source

For foods with low to moderate carbohydrate density, a 50 g available carbohydrate load may require ingestion of large volumes.

“However, in the case of foods with low to moderate CHO density, this may lead to large volumes having to be ingested.”

The fasted condition is the most stable with respect to possible intra-individual differences due to time of day and meal influences.

“This fasted condition is most stable concerning possible intra-individual differences due to time of the day and meal influences.”

The difference in glycaemic response between two breakfast cereals tested in the morning was significantly greater than the difference observed at lunchtime.

“It was observed that the difference in glycaemic response between the cereals when tested in the morning was significantly greater than that observed at lunchtime (Wolever & Bolognesi, 1996c).”

• cites Wolever & Bolognesi, 1996cNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The validity of estimates of GI of available carbohydrate and glycaemic impact remains to be better established.

“In general the validity of the above estimates of GI of available CHO and glycaemic impact remains to be better established.”

Evening meals with a low GI produce better glucose tolerance the following morning than evening meals with a high GI.

“Evening meals of low GI produce better glucose tolerance the following morning compared with evening meals of high GI (Wolever et al. 1988b;Thorburn et al. 1993).”

• cites Wolever et al. 1988bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Thorburn et al. 1993 sourceNot checked — no full text of the source

Fingertip capillary blood appears to give the greatest sensitivity.

“Fingertip capillary blood appears to give the greatest sensitivity.”

The considerations of randomisation and allocation concealment from parallel-design drug trials are difficult or impossible to apply to GI studies.

“These considerations apply to parallel-design clinical trials of drugs, but are difficult or impossible to apply to studies in which the GI of foods is determined.”

• AssumptionRests on an assumption: GI trials are cross-over studies in which all subjects undergo all treatments.
• AssumptionRests on an assumption: It is usually impossible to blind either subjects or investigators in GI trials.

Because tissues consume glucose, peripheral venous glucose concentration is lower than arterial.

“Because the tissues consume glucose, the concentration of glucose in a peripheral vein will be lower than that in an artery.”

Certain foods produce higher insulin responses than expected from their GI.

“It should be acknowledged though that certain foods produce higher insulin responses than expected from their GI (Ostman et al. 2001).”

• cites Ostman et al. 2001 sourceNot checked — no full text of the source

For legumes, RGR markedly increased as length of blood sampling increased.

“For legumes, the RGR markedly increased as the length of time of blood sampling increased.”

• cites Gannon & Nuttal (1987) sourceNot checked — no full text of the source

A washout period of 2 days seems to be enough.

“But it seems that a washout period of 2 d is enough.”

• AssumptionRests on an assumption: The impact of the food tested on metabolism 2 days later is negligible.

Methods that result in GI values that do not control for differences in AUC between individuals cannot be considered valid.

“Therefore, methods that result in GI values that do not control for differences in AUC between individuals cannot be considered to be valid.”

• AssumptionRests on an assumption: GI is a property of the food independent of subject characteristics.

Using plasma and whole blood inconsistently in a series of measurements would be inappropriate.

“However, it would not be appropriate to use plasma on one occasion and whole blood on another.”

For starchy foods such as bread, oats and potato, there was only a small effect of time of blood sampling on RGR.

“For starchy foods such as bread, oats and potato, there was only a small effect of time of blood sampling on RGR.”

• cites Gannon & Nuttal (1987) sourceNot checked — no full text of the source

Mean and variation of GI values tend to increase as frequency and duration of blood sampling decreases.

“The mean and variation of the resulting GI values also tend to increase as the frequency and duration of blood sampling decreases.”

• cites Wolever, 2004bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The site of capillary blood sampling may affect results.

“The site at which capillary blood is taken may affect the result.”

After sucrose, milk and fructose, RGR markedly decreased as length of blood sampling increased.

“After sucrose, milk and fructose, the RGR markedly decreased as the length of time over which blood was sampled increased, presumably because blood glucose concentrations after these test meals undershot more than after glucose.”

• cites Gannon & Nuttal (1987) sourceNot checked — no full text of the source

Since area above fasting glucose is only a small proportion of total AUC, total AUC is insensitive for distinguishing foods with different glucose-raising effects.

“Since the area above the fasting glucose is only a small proportion of the total AUC, total AUC is insensitive for distinguishing between foods with different glucose-raising effects.”

Fingertip capillary sampling is preferable to other sites.

“The message is that fingertip capillary sampling is preferable to other sites.”

Adhering to a standardised schedule of blood sampling with respect to frequency and length is likely important for valid and consistent GI values.

“Thus, adhering to a standardised schedule of blood sampling with respect to both frequency and length of time is likely to be important for obtaining valid and consistent GI values.”

Use of different methods of calculating AUC results in different GI values.

“Since use of different methods of calculating AUC results in different GI values (Tables 2 and 3), one argument for retaining this method is that adoption of a new method might render all previous GI values inaccurate.”

• cites Tables 2 and 3Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

In one study, antecubital venous glucose was 4 mmol/l lower than arterialised venous blood at 60 min after glucose ingestion.

“In one study, the glucose concentration in blood taken from an antecubital vein was 4 mmol/l lower than that in arterialised venous blood at 60 min after ingestion of glucose (Frayn et al. 1989).”

• cites Frayn et al. 1989 sourceNot checked — no full text of the source

Venous glucose measurement might be expected to be more variable than capillary.

“Therefore, venous glucose measurement might be expected to be more variable than capillary.”

The test meal can only affect AUC above the fasting level, unless there is an undershoot.

“The test meal can only affect AUC above the fasting level (unless there is an undershoot).”

• AssumptionRests on an assumption: Fasting glucose is independent of the food consumed.
• AssumptionRests on an assumption: The portion of total AUC below fasting glucose is independent of the test meal.

GI measurements were not affected by capillary versus venous blood because reference and test measurements were similarly affected.

“However, the measurements of GI were not affected as both reference and test measurements were similarly affected.”

• cites Granfeldt et al. 1995bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Meals eaten between sessions have more effect on the next test than the test food eaten 2 days before.

“Moreover, the subject eats between the two sessions, and these meals have more effect on the next test than the portion test eaten 2 d before.”

• AssumptionRests on an assumption: Meals between sessions have more effect on the next test than the test food eaten 2 days before.

Sensitivity of measurement was greater using capillary blood sampling.

“Sensitivity of the measurement, however, was greater using capillary blood sampling (Granfeldt et al. 1995b).”

• cites Granfeldt et al. 1995bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Making two test sessions on successive days or with a 1-day washout may not be enough.

“Thus, making two test sessions on two successive days, or with a washout period of 1 d, may not be enough.”

Reduced frequency and duration of blood sampling resulted in statistically significant correlations between mean AUC of reference food and mean GI of test foods.

“'In addition, reduced frequency and duration of blood sampling resulted in the development of statistically significant correlations between the mean AUC of the reference food and the mean GI of the test foods in the different subjects, a correlation which is not significant' (Wolever, 2004b).”

• cites Wolever, 2004bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Taking blood samples less frequently or for less than 2 h in normal subjects significantly influences mean and variation of AUC.

“Taking blood samples less frequently (sampling blood every 30 min instead of every 15 min) or for less than 2 h in normal subjects significantly influences the mean and variation of the resulting AUC (Wolever, 2004b).”

• cites Wolever, 2004bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The arterio-venous glucose difference may be considerable, especially after glucose ingestion when insulin stimulates glucose uptake.

“The magnitude of this arterio-venous difference may be considerable, especially after glucose ingestion when insulin stimulates glucose uptake.”

The international GI table entries for oatmeal are for raw oat bran, not oat meal.

“The value used for oatmeal, entries 116 and 117, are not for oat meal but for raw oat bran (high-fibre extract of oats), etc.”

Available evidence indicates that GI values of composite high-carbohydrate meals can be predicted well from the meal components.

“Instead, available evidence indicates that GI values of composite high-CHO meals can be predicted well based on the meal components.”

Outliers can have considerable influence on the results of a statistical analysis.

“Outliers are particularly important because they can have considerable influence on the results of a statistical analysis (Altman, 1991).”

• cites Altman, 1991Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

When standardised methods are applied to determine GI (glucose=100) of identical high-carbohydrate foods, the interlaboratory SD of GI values is approximately 9.

“When standardised methods are applied to determining the GI (glucose ¼ 100) of identical high-CHO foods, the interlaboratory SD of GI values is approximately 9.”

• AssumptionRests on an assumption: The five foods and five laboratories in the interlaboratory study are representative enough that the pooled SD of about 9 applies to standardised GI measurements generally.

The probability of such differences between any two of a group of laboratories is higher than the calculated probability based on exactly two laboratories.

“It should be noted that these calculations are based on exactly two laboratories.The probability of such differences between any two of a group of laboratories will be higher.”

The quality and quantity of carbohydrate determine the glucose response to a meal.

“However, it remains the quality and the quantity of CHO that determine the glucose response to the meal (Wolever & Bolognesi, 1996b).”

• cites Wolever & Bolognesi, 1996bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Because fat and protein are known to influence glucose and insulin responses, GI of meal components alone would not be expected to predict glycaemic responses when fat and protein vary widely.

“In the face of such large differences in fat and protein, nutrients, which are known to influence glucose and insulin responses (Nuttall et al. 1985;Ercan et al. 1994), it is expected that GI of the meal components alone would not predict glycaemic responses.”

• cites Nuttall et al. 1985 sourceNot checked — no full text of the source
• cites Ercan et al. 1994 sourceNot checked — no full text of the source

Other macronutrients (fat, protein, carbohydrate, and other nutrients) from co-ingested foods influence both glucose and insulin responses.

“Indeed, the other macronutrients (fat, protein, CHO and other nutrients) provided by the co-ingested foods have an influence on both glucose and insulin responses.”

Glucose estimation methods with a CV greater than 3% will introduce significant analytical variation into GI measurement.

“Methods for glucose estimation with a CV .3 % will introduce significant analytical variation into the measurement of GI.”

When a food is included in a meal, the glucose response to the meal is not the same as the glucose response to the single food.

“When a food is included into a meal, the glucose response per se to this meal will not be the same as the glucose response to the single food.”

If interlaboratory SD decreases to 8, 7, 6, or 5, the respective 95% difference falls to 23, 20, 17, and 14.

“If the interlaboratory SD were decreased to 8, 7, 6, or 5, the respective difference would fall to 23, 20, 17 and 14.”

An interlaboratory SD below 10 would likely be acceptable because it would mean the 95% CI for any GI value would be less than about ±20.

“An interlaboratory SD of , 10 would likely be acceptable because this would mean that the 95 % CI for any GI value would be , ^20.”

The GI of a single food depends on its composition.

“First of all, the GI of a single food is dependent on the composition of this food.”

The GI concept remains discriminating in the context of a meal.

“This means that the GI concept remains discriminating in the context of a meal (Bornet et al. 1987).”

• cites Bornet et al. 1987 sourceNot checked — no full text of the source

The mean of ratios and the ratio of means do not yield exactly the same result when applied to the same data.

“Using the same set of data, these methods do not yield exactly the same result (Wolever et al. 2003).”

• cites Wolever et al. 2003Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

With an interlaboratory SD of 9, there is a 95% chance that the difference in GI value for the same food in two different laboratories is less than 25.

“With an interlaboratory SD of 9, there is a 95 % chance that the difference in GI value obtained for the same food in two different laboratories is , 25.”

Indexing individual glycaemic response to a standard minimises between-individual variation due to subject and food-related factors.

“The advantage in dealing with GI values is that by indexing the individual glycaemic response to a standard, the variation that occurs due to several factors, for example, subject and food-related ones, the between-individual variation is minimised (Wolever, 1990a).”

• cites Wolever, 1990aNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

With an interlaboratory SD of 9, the chance that two laboratories obtain GI values for the same food differing by less than 15 is about 76%.

“With an interlaboratory SD of 9, the chance of two laboratories obtaining a GI value for the same food which differs by less than 15 is about 76 % (i.e. a 24 % chance that the difference is greater than 15).”

The lack of correlation in Flint et al. 2004 is not unexpected because the test meals had large ranges of fat, protein, and energy despite containing 50 g available carbohydrate.

“These results are not unexpected because, although the test meals contained 50 g available CHO, there was a large range of fat (3 -42 g), protein (5 -28 g) and energy (1130 -2990 kJ).”

In some subjects, foods with readily available carbohydrate such as white bread may cause a sharp rapid decline in blood glucose within 120 min and late postprandial hypoglycaemia.

“In certain subjects, foods with readily available CHO, such as white bread, may induce a sharp and rapid decline in blood glucose within the 120 min period, and a hypoglycaemia in the late postprandial phase.”

In an interlaboratory study with five laboratories using capillary blood sampling to determine GI (glucose=100) of four centrally provided foods and locally obtained white bread, the pooled interlaboratory SD of GI values for the five foods was approximately 9.

“The pooled interlaboratory SD of the GI values for the five foods was approximately 9.”

• cites Wolever et al. 2003Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• AssumptionRests on an assumption: The five foods and five laboratories in the interlaboratory study are representative enough that the pooled SD of about 9 applies to standardised GI measurements generally.

Factors linked to the quality and quantity of carbohydrate in a food can also affect its GI.

“Moreover, other factors linked to the quality and the quantity of CHO present in the food can also affect its GI (see also Table 1).”

Using glucose as the reference for GI is not a good method because the resulting GI values depend on the subject's glucose tolerance status.

“glucose), suggesting that this is not a good method of determining GI, since the GI values obtained depend on the glucose tolerance status of the subject.”