An insulin index of food

Executive summary

* The Holt et al classic paper, published in 1997, reviewed the insulin (and glucose) responses of 38 different foods.

* The 38 foods were placed in 6 food categories: fruit; bakery products; snack foods and confectionery; protein-rich foods; carbohydrate-rich foods; and breakfast cereals. (There was a conflict of interest with one of the authors and Kellogg’s.)

* The researchers should have tested single macronutrients. Every food other than white fish was a composite of at least two macronutrients and mostly all three. The fish approximated to pure protein, but sucrose and an oil could have been tested as a pure carbohydrate and fat respectively. This was a missed opportunity in a classic study.

* Healthy subjects tested each category of foods (11-13 students tested the foods in each category). The areas under the glucose and insulin curves were then measured for 120 minutes after the single food. The researchers aimed to replicated the glycemic index by assigning white bread a score of 100 and measuring the other 37 foods relative to this.

* Three conclusions (among others) were i) that there were significant differences in insulin scores within and between food categories; ii) that protein had surprising responses; and iii) that breakfast cereals were the lowest category for average insulin scores and the second lowest for average glucose scores.

* The categories were arbitrary and wrongly influenced the outcomes. I placed the foods in different categories – real or processed foods and then mostly fat, carbohydrate or protein foods. The findings changed and beef, fish, peanuts, cheese and eggs became the 'best' foods and cereals were among the worst.

* The glucose and insulin impact of protein was interesting and is explored in this note.