Glycemic index and glycemic load of foods

The index answers “how fast”, the load answers “and how much”. The index is measured on a serving containing exactly 50 g of available carbohydrate: for watermelon that is over 600 g of flesh, for white bread about 100 g. Comparing by index therefore compares these artificially equalised servings rather than the ones people eat. Load puts the portion back: GI × carbohydrate in the portion ÷ 100. A practical rule: use the index to choose between two side dishes, and the load to understand what a particular plate will do.

Because there is almost no carbohydrate in it: around 8 g per 100 g of flesh, the rest is water. A GI of 76 means the carbohydrate that is there is absorbed fast. But reaching the reference 50 g would take over half a kilo. A realistic 250 g slice carries roughly 20 g, giving a load around 15 — medium. Exactly the same mechanism applies to pumpkin, beetroot and carrots: all of them spent years on “forbidden” lists because of an index, while the load of a portion is modest.

The reference used in research: up to 80 per day is low, above 120 is high. For a single meal, up to 10 is low, 11–19 medium, 20 and above high. These thresholds are a frame for comparison, not a medical target: someone with diabetes gets their targets from a doctor and should follow meter readings, not a table. For everyone else the numbers are useful in another way — they show that the biggest contributor is usually an ordinary side dish rather than the dessert it is supposed to replace.

No, and it is the most common mistake made with these tables. The index is a ratio, not a quantity: an average between bread and butter means nothing, because butter has no carbohydrate at all. Only loads add up, because each of them has already been scaled to a real portion. That is what the plate here is for: it sums loads, carbohydrate and calories, and it usually turns out that nearly all of a meal’s load came from a single line rather than being spread across all of them.

Because the spread is real and often enormous. Published values for boiled potato run from 56 to 101 depending on variety, tuber size, boiling time and whether it was eaten hot. Pasta runs from 32 to 78. The international tables publish both the mean and the range, but websites usually print only the mean — so two sources appear to contradict each other while both took their numbers from the same paper. Here the range sits next to the value: when it is wide, that width is the most important thing to know about the food.

Four tricks with a measured effect. First, cool cooked starch — part of it turns resistant, and the response to cold or reheated potato, rice and pasta drops by 10–25 %. Second, do not overcook: al dente keeps the index roughly 20 points below mushy. Third, acid — vinegar or lemon juice slows gastric emptying and cuts the response by 15–30 %. Fourth, order — starting with vegetables and protein and eating carbohydrate last visibly lowers the peak from the very same plate. None of them reduces calories, and none of them cancels the amount of carbohydrate.

The glycemic index answers how far blood sugar rises; the insulin index answers how much insulin is released in response. They usually travel together but come apart for dairy and protein foods: fruit yogurt has an index of 41 and an insulin response of 115, higher than white bread, while meat, fish, eggs and cheese have no glycemic index at all — they contain no carbohydrate — and still raise insulin. The two numbers cannot be compared directly: the insulin index is measured on portions of equal energy (1000 kJ), the glycemic index on portions containing 50 g of carbohydrate. Published values are scarce: Holt’s original study covered 38 foods, of which 26 match this table. An empty cell means "not measured", not "zero".