Diabetes 2013 May (2)-1


Is there a dose-response relation of dietary glycemic load to risk of type 2 diabetes? Meta-analysis of prospective cohort studies.

Geoffrey Livesey, Richard Taylor, Helen Livesey, and Simin Liu. Independent Nutrition Logic, Wymondham, Norfolk, UK. glivesey@inlogic.co.uk

Am J Clin Nutr.2013 Mar;97(3):584-596.


Background: Although much is known about the association between dietary glycemic load (GL) and type 2 diabetes (T2D), prospective cohort studies have not consistently shown a positive dose-response relation.

Objective: We performed a comprehensive examination of evidence on the dose response that links GL to T2D and sources of heterogeneity among all prospective cohort studies on healthy adults available in the literature.

Design: We conducted a systematic review of all prospective cohort studies and meta-analyses to quantify the GL-T2D relation both without and with adjustment for covariates.

Results: Among 24 prospective cohort studies identified by August 2012, the GL ranged from ~60 to ~280 g per daily intake of 2000 kcal (8.4 MJ). In a fully adjusted meta-analysis model, the GL was positively associated with RR of T2D of 1.45 (95 % CI: 1.31, 1.61) for a 100-g increment in GL (P<0.001; n = 24 studies; 7.5 million person-years of follow-up). Sex (P = 0.03), dietary instrument validity (P<0.001), and ethnicity (European American compared with other; P = 0.04) together explained 97 % of the heterogeneity among studies. After adjustment for heterogeneities, we used both funnel and trim-and-fill analyses to identify a negligible publication bias. Multiple influence, cumulative, and forecast analyses indicated that the GL-T2D relation tended to have reached stability and to have been underestimated. The relation was apparent at all doses of GL investigated, although it was statistically significant only at above 95 g GL/2000 kcal.

Conclusion: After we accounted for several sources of heterogeneity, findings from prospective cohort studies that related the GL to T2D appear robust and consistently indicate strong and significantly lower T2D risk in persons who consume lower-GL diets. This review was registered at http://www.crd.york.ac.uk/PROSPERO as CRD42011001810. Am J Clin Nutr doi: 10.3945/ajcn.112. 041467.



Glycemic load (GL) is the product of carbohydrate amount in a food item and the item’s Glycemic index (GI). This index represents the ability of a food carbohydrate once eaten to elevate the concentration of glucose circulating in blood relative to the elevation from a similar amount of glucose taken in water. Hence an increase in either carbohydrate intake or GI of the carbohydrate food consumed increases a person’s GL so exposing them to a higher concentration of blood glucose (Glycemic response), which risks damage to tissues in the body.

Geoffrey Livesey-1We examined the relation between GL and the incidence of type 2 diabetes (T2D) for 7.5 million person-years of follow-up among 24 prospective cohort studies of ostensibly healthy adult persons from any part of the globe that we had identified by systematic searches of the literature published before August 2012 (Figure; reproduced from the Am J Clin Nutr doi: 10.3945/ajcn.112. 041467 Supplemental Data file). Bubbles in the figure each represent the result of an individual study and are of different sizes—the more precise studies have larger bubbles to represent the greater weight of evidence. The figure shows the relative risk (RR) of developing T2D was higher (RR>1) among sub-population cohorts of persons consuming diets of higher GL, and that the risk varied with the sex of the sampled population. The relative risk shown in the figure is for a 100 g increment in GL in 2000 kcal of diet.

Follow-up periods were from 4 to 26 years and the strength of association was maintained throughout all periods.

The central fitted line in the figure shows a trend towards a higher risk among sampled females than among males, and intermediate results for population samples of mixed sexes. The curved lines either side of the trend line show the trend’s 95% confidence intervals. That the lowest of these lines fails to fall to RR=1 at its lowest point indicates a statistically significant GL-T2D relation existed both in males and in females.

Deviations in relative risks from the trend across sexes were explained by study imprecisions (e.g. small studies) since the residual deviations fitted a symmetrical funnel plot.

The observations shown in the figure were adjusted for centered covariates that represented the ethnicity of participants and the validity of the dietary instruments used to assess GL consumption. These covariates were statistically significant as was the trend shown in the figure for the sex of participants in the sampled populations. Together the covariates explained 97 % of the statistically real inconsistency (heterogeneity) in results among studies. Prior absence of explanation of heterogeneity had caused some doubt among authorities considering the utility of GL (or GI) in health maintenance. This explanation of the heterogeneity adds to the view that dietary control of glycemia has real biological and potentially economic importance.

Analysis showed that the GL-T2D relation was underestimated by regional studies compared with the larger world-wide (global) variation in GL among individuals, and was shown to be underestimated also in all studies due to imperfect validity of the dietary instruments used to assess GL.

No threshold on GL in a 2000 kcal diet was identified below which there was no GL-T2D relation. The relation appeared continuous and log-linear across all increments for GL from the lowest to the highest loads consumed world-wide. However, there was insufficient evidence to indicate statistical significance for this relation below a GL of 95 g in each 2000 kcal of diet consumed. This was close enough to 100 to propose 100 g as an evidently optimal rounded target for minimizing the risk of developing T2D among ostensibly healthy persons.

In our article we explain how this target might be reached by consuming foods of lower GI, but also that persons consuming the highest amounts of carbohydrate would also have to limit carbohydrate intake as well as choosing to eat lower GI foods.

In a previous publication [1] it was noted that choosing to eat a higher proportion of foods that are wholegrain, fruit, or vegetables, can help to lower the GI and hence GL from carbohydrate foods. But, it was shown also that foods in each of these food categories vary considerably in their GI, such that diets optimal for reducing the risk of T2D could not be expected by such food-based advice alone. Meanwhile, in a meta-analysis of intervention studies [2], diets of lower GI or GL were shown to improve glycemic control and glycated proteins in T2D patients independently of their dietary fiber content.


1. Livesey, G., Glycemic response and toleration, in Sweeteners and Sugar Alternatives in Food Technology, 2e, K. O’Donnell and M.W. Kearsley, Editors. 2012, Wiley-Blackwell: Oxford. p. 1-26.

2. Livesey, G., et al., Glycemic response and health a systematic review and meta-analysis: relations between dietary glycemic properties and health outcomes. Am J Clin Nutr., 2008. 87(1): p. 258S-68S.


Geoffrey Livesey-2Geoffrey Livesey1  was lead author in this study and thanks his co-authors Helen Livesey1, Richard Taylor2, and Simin Liu3.  He is Director of a consultancy firm Independent Nutrition Logic Ltd, and can be contacted via email at glivesey@inlogic.co.uk.

1Independent Nutrition Logic, Wymondham, Norfolk, UK.  2Merton College, Oxford, UK. 3University of California Los Angeles, USA.

Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann