The founding measurements, including the two things everyone assumes and the study disproved.
What the evidence says
- The index was built by testing *62 commonly eaten foods on groups of 5–10 healthy fasting volunteers*, measuring the 2-hour blood-glucose area under the curve as a percentage of the equivalent glucose load[1]
- *Food-group means, ascending:* legumes 31 ± 3% · dairy 35 ± 1% · fruit 50 ± 5% · biscuits 60 ± 3% · cereals 60 ± 3% · breakfast cereals 65 ± 5% · root vegetables 72 ± 6%[1]
- Legumes came in dramatically lower than cereals — mean 31% against 59% (p < 0.001)[1]
- *Fibre did NOT predict glycemic response (r = −0.220, not significant), and neither did sugar content (r = 0.022, NS)*[1]
- *The bread comparison makes the same point concretely: wholemeal bread (72 ± 6%) was not lower than white (69 ± 5%).* Wheat fibre did not reduce the glycemic response[1]
- *What did predict it was fat and protein* — fat negatively correlated (r = −0.386, p < 0.01) and protein more strongly still (r = −0.523, p < 0.001)[1]
- *Food form outweighed fibre content:* wholemeal spaghetti (42 ± 4%) came in at roughly half the GI of wholemeal bread (72 ± 6%) — same grain, same fibre, different physical structure[1]
- Fructose scored only 20 ± 5%, which is why sucrose was moderate at 59 ± 10% despite being a "sugar"[1]
- Dose-response curves for glucose, bread and lentils showed GI differences were *accentuated rather than diminished* at higher carbohydrate loads (>50 g)[1]
- The results correlated with the only prior comparable study at r = 0.823 (p < 0.001) across 10 common foods[1]