Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • ADVERTISEMENT FEATURE Advertiser retains sole responsibility for the content of this article

Is it time to rethink the glycemic index?

A cooled potato salad has a different GI score than the same meal at a higher temperature.Credit: Caia Image/Getty Images

Over the past half-century, type 2 diabetes cases in the United States have risen precipitously. Just 1% of adults had the disease in 1958; by 2020 it was 13% — around 34 million people1.

“It’s a very serious public health problem, and we need to address it,” says Mindy Patterson, an Associate Professor of Nutrition at Texas Woman’s University, and dietitian.

One of the key tools that physicians draw on to manage or prevent diabetes in patients is the glycemic index (GI), which ranks the impact of carbohydrate-containing foods on blood glucose levels on a scale of one to 100. Since it was developed in 1981, awareness of the index has spread beyond clinicians. Many people now consider it to be a metric for healthy nutrition.

However, use of this 40-year-old paradigm has become controversial among nutrition scientists and medical practitioners, says Joanne Slavin, a dietitian and researcher at the University of Minnesota’s Department of Food Science and Nutrition. “The index ignores many factors that determine how quickly carbohydrates are digested and absorbed,” she explains. Key omissions include how foods are grown, manufactured, cooked and stored; serving size; and the combination of foods eaten as part of a meal2.

The GI also doesn’t account for substantial heterogeneity in blood-glucose responses to the same food by different people, or even by the same person on different occasions, Slavin adds.

These problems have incentivized researchers to search for more comprehensive ways to evaluate carbohydrates and personalize glycemic control — while incorporating other nutritional indicators such as fibre content. The goal is to help health-care professionals support their diabetes patients to choose the most appropriate foods. And these foods, Patterson adds, should be affordable, available, and respectful of cultural eating patterns so that recommendations work in the real world.

Carbs are complex

The GI, with its narrow focus on glucose, skews the health value of some carbohydrate foods, says Slavin. Fructose, she points out, ranks a healthy-sounding 19 on the GI, but its consumption is associated with diabetes and heart disease3 — another oversight because the GI does not consider chronic disease risk. A candy bar, which is loaded with sucrose but contains little glucose, gets a better GI score than a slice of whole-grain bread, she adds.

Furthermore, there are numerous rankings for the same food, making the GI confusing for physicians and patients. Patterson cites the example of rice, which has 126 different entries for various brands and regional varieties, further sub-divided by cooking method4. The GI scores range from 37 (for Chinese rice vermicelli, cooked for 8 min) to 116 (Jasmine rice, Reindeer brand, made in a rice cooker).

These differences can be compounded by how the GI score is determined. When seven different labs assessed the same type of rice, GI scores ranged from 55 to 875.

Serving temperature also affects a food’s score. Cooked rice, when eaten hot, causes a higher glycemic response than the same rice eaten cold6. The same is true of other starchy foods like potatoes. “Cooking and cooling starch alters the amylose content,” says Patterson. This process increases the content of ‘resistant starch’ in carbohydrates, which then acts more like dietary fibre — moderating the glycemic response, she says.

One of the biggest issues with the index, says Patterson, is that it tests carbohydrates alone to derive a GI score, but these foods are rarely eaten in isolation. Most are consumed as part of meals also containing fat, protein and fibre, which alter glucose absorption and blood-sugar levels. “Butter the bread,” she says, “and the glycemic index drops.”

Attempts have been made to modify the GI to better reflect real life. The glycemic load (GL) adjusts GI scores for portion sizes, but it creates some confusing scenarios. High-GI foods may have a low GL. Watermelon, for example, has a GI of 72, but the GL of a standard-sized portion is only 4. What’s more, the calculations are complicated: to find the glycemic load, the GI must be multiplied by the amount of carbohydrate in a serving size, then divided by 100. Not only does the GL suffer from the same issues as the GI, it is impractical for real world situations.

It’s not one size fits all

Ongoing research continues to highlight individualized glycemic responses to foods. One study monitored 800 people for a week, finding a five-fold difference in post-meal glucose levels between the top and bottom 10%, though all had eaten the same foods. Dietary habits, physical activity, body composition and gut microbiota were all important factors in response7.

A more recent study assessed the postprandial metabolic responses of more than 1,000 healthy adults, eating identical meals over a 2-week period. Continuous glucose monitors mapped blood sugar levels and revealed large inter-individual variability, even between identical twins — suggesting that genetics were not the main cause of the differences8.

Such is the variability that even the same person can exhibit different responses to the same food eaten on consecutive days9 or at different times of the day10.

Altogether, such studies have found that many biological and behavioural factors influence glycemic response, including age, stress, health status, baseline insulin levels, alcohol consumption and sleep.

The upshot is that people with type 2 diabetes, and their physicians are in need of more holistic dietary tools. “Current approaches that demonize staple carbohydrate foods do little to promote the recommended pattern of foods known to improve health status and reduce disease risk,” Slavin wrote in one recent analysis11.

In 2021, an expert panel report outlined ways to revamp carbohydrate assessment12. The authors — including Slavin — outlined the need for a more expansive, standardized, evidence-based way to evaluate carbohydrate quality.

The panel envisioned a single simple carbohydrate-food metric that melds multiple intrinsic properties, including total fibre content, nutrient density, food group designation and processing effects. While such a metric will not fully capture all of a food’s health effects, the panel believes that it would make the metric more realistic and better able to inform public health policies, including the 2025–2030 Dietary Guidelines for Americans.

By separating the intrinsic food values from the extrinsic effects, there is a clearer pathway to fully personalized nutrition, says Patterson. Research can then build by better understanding the effects that carbohydrate quality has on human health. Many labs are already working towards developing such predictive models.

A more expansive analysis of healthy carbohydrates that ends the confusion over ‘high-GI’ plant-based, fibre-rich foods will give people, particularly those with diabetes, better information to make dietary decisions.

For more information about this issue, check out some of the latest science and resources at


  1. CDC’s Division of Diabetes Translation. United States Diabetes Surveillance System April 2017.

    Google Scholar 

  2. Naser, K.A. & Wimalawansa, S.J. SRL Diabetes Metab. 1, 001-004 (2015).

    Google Scholar 

  3. Andres-Hernando, A. et al. Cell Metab. 32, 117–127 (2020).

    Google Scholar 

  4. Atkinson, F., Brand-Miller, J., Foster-Powell, K., Buyken, A.E. & Goletzke, J. Am J Clin Nutrition 114(5):1625–1632 (2021).

    Google Scholar 

  5. Wolever, T. Am J Clin Nutr. 106, 704–705 (2017).

    Google Scholar 

  6. Dhar, Amrit, et al. Int J Res Med Sci, 9, 828-832 (2021).

    Google Scholar 

  7. Zeevi, D. et al. Cell 163, 1079–1094 (2015).

    Google Scholar 

  8. Berry, S.E., et al. Nature Medicine 26, 964–973 (2020).

    Google Scholar 

  9. Matthan, N.R. et al. Am J Clin Nutrition 104(4):1004 (2016).

    Google Scholar 

  10. Devlin, B., Parr, E., Radford, B. & Hawley, J. Clin Nutrition 40, 2200–2209 (2020).

    Google Scholar 

  11. Schulz, R. & Slavin, J. Adv Nutr 12, 1108–1121 (2021).

    Google Scholar 

  12. Comerford, K.B., et al. Nutrients 13(8):2667 (2021).

    Google Scholar 

Download references


Quick links