What is the Glycemic Index?
The glycemic index (GI) was developed by Dr David Jenkins at the University of Toronto back in the early 80s. The GI of a food was determined by having study subjects fast for 12 hours before consuming a specific food containing 50 grams of carbohydrates.
The GI “score” attributed to each food was determined by the area under the blood glucose curve over a 2 hour period after food ingestion – I’ve drawn an example below.
It was originally intended to provide a measure of which carbohydrates would be more or less suited to people with type 2 diabetes or impaired glucose tolerance – the hypothesis being that lower GI foods would be superior for diabetics.
There are numerous flaws when it comes to the application of the glycemic index:
- It only applies when you eat the food in isolation –anything from the list below can alter the GI of a food:
- Adding starch, fibre, or sugars
- Adding fat or protein
- Adding organic acids (such as vinegar) or salt
- Homogenization – i.e. combinations of fluids that aren’t usually able to be mixed
- Chewing food for longer
- Eating servings of carbohydrates of more or less than 50 grams will alter your blood glucose response compared to that predicted by the GI
The American Diabetes Association acknowledges that GI should not be used as one of the foundations of a diet – they recommend that some form of carbohydrate counting should form the basis, with GI being used to “fine tune” your intake. However, it’s still the basis of a lot of popular diets, and many people still believe that it plays an important role in body composition.
Enrolment for the SBS Academy opens next on February 28th, 2018. Hit the button if you have any questions.
The Final Nail In The GI Coffin?
In November of 2015, a paper was published in a journal called Cell by some researchers at the Weizmann Institute of Science by David Zeevi and colleagues – “Personalized Nutrition by Prediction of Glycemic Responses”.
The researchers wanted to ultimately design an algorithm to help adapt peoples’ diets to better control their blood glucose levels, given that prediabetes and type 2 diabetes are both on the rise.
The picture above shows an overview of the study. The first phase involved the monitoring of 800 non-diabetic members of the Israeli population – with obesity levels at 22%, and 54% of the sample being overweight (BMI > 25), these are fairly representative of the current levels of overweight and obese people in other developed countries.
In these 800 people, the researchers spent a week having the participants log everything – activity, food (from a database developed by the researchers) and sleep – in real time. The participants were also given subcutaneous glucose monitors, which are incredibly accurate and monitor blood glucose levels every 5 minutes. To increase the accuracy of dietary reporting, the participants were all informed of the importance of accurate reporting in order to obtain accurate glucose levels. During this week, the participants were fed one of 4 standardised meals (bread, bread and butter, glucose or fructose) as breakfast, containing 50g of available carbohydrates per meal.
The researchers also analysed each participant’s gut bacteria profile (via stool samples – being a researcher is so glamorous) to see if they could correlate levels of any bacteria with glucose responses.
The researchers then developed an algorithm based on the data to predict an individual’s glucose response to food, before testing it in the “validation cohort” of 100 participants.
Finally, another algorithm was developed to help put in place dietary interventions to minimise blood glucose response. This was tested in 26 participants: 12 ate a diet designed by the algorithm, and 14 ate a diet designed by an expert who specialised in diabetic nutrition.
What did the study find?
Post-Prandial Glucose Response (PPGR)
What the results showed was fairly startling – the differences in blood glucose response to the various foods was hugely different from participant to participant. The figures below show the extent to which this happened.
The PPGR of an individual was also correlated with the presence of certain gut bacteria – Proteobacteria and Enterobacteriaceae, specifically.
What does this show? In my opinion, this provides pretty irrefutable evidence that the Glycemic Index alone is largely irrelevant when attempting to control blood glucose response to meals.
The Prediction Algorithm
The question then arises – can we predict PPGR? If so, what factors are actually at play when it comes to predicting someone’s PPGR?
The authors developed a method of doing this, and with the validation cohort of 100 people showed that it predicted PPGR fairly well for a given meal (the correlation coefficient, R, which usually falls between 0 and 1, was around 0.7). The researchers incorporated the following factors into their algorithm:
- The carbohydrate content of the meal – for ~95% of the participants, more carbohydrates correlated with an increased PPGR. However, there were two distinct categories:
- People who were “carbohydrate sensitive”, and showed a much larger PPGR after carbohydrate ingestion
- People who were “carbohydrate insensitive”, and showed a much smaller increase after consuming the same quantity of carbohydrates
- The fat-to-carbohydrate ratio of the meal – some studies have previously shown that adding fat to a meal reduces the PPGR, and this study confirms that… but only in some people. In others, adding fat to a meal had little-to-no effect on PPGR compared to the carbohydrate content of the meal. So we have “fat sensitive” and “fat insensitive” people now, too!
- Fibre – whilst increasing the fibre content of a meal seemed to increase the PPGR (somewhat counter-intuitively), the amount of fibre ingested in the 24 hours previously had a PPGR-lowering effect.
- Sodium, time from last sleep, cholesterol levels and age all exhibited negative effects on PPGR (as in – they go up, PPGR goes up).
- Meal water and alcohol content display positive effects on PPGR (they go up, PPGR goes down).
- Gut bacteria profile – this was complicated; certain bacteria showed positive, negative and neutral effects on PPGR.
The PPGR-Minimizing Algorithm
The researchers’ algorithm to minimise PPGR was just as good (if not slightly better) as a diet designed by an expert – which is really exciting for the development of future nutrition guidelines to help treat diabetes and obesity. It obviously needs to be tested in a much larger sample size than 12 people, and needs a lot more development before it could be implemented into a treatment plan; however, it starts to set the scene for what could happen with personalised nutrition in the future.
GI – Dead, Buried and Irrelevant
In conclusion – the glycemic index is a completely outdated metric on which to base your diet or your carbohydrate choices. There appears to be no correlation between the GI of a food and an individual’s PPGR; as a result, trying to design a diet around the GI of individual foods is about as useful as trying to programme your training based on the length of your little toe.
An individual’s response to food is likely to be highly individualised. This needs to be recognised and accommodated for when it comes to nutrition recommendations – whether those be for the prevention of diabetes, reducing obesity, or sporting performance.