Amylose content outperforms digestion kinetics in predicting the glycemic index of cooked rice

This study developed and compared two oral-phase pretreatment strategies, mastication (physiologically realistic) and homogenization (standardized), for in vitro digestion of freshly cooked rice using a combined parallel-sequential kinetics model. Univariate GI prediction models based on AC, rapidly digestible starch weight ( [Formula: see text] ), and 120-min area under the digestion curve (AUC₁₂₀) were developed from seven rice varieties with contrasting amylose content (AC, 8.1% ~ 31.9%), and were externally validated on four varieties from a separate same-laboratory experiment, while an additional nine varieties were used for extended prediction. The homogenization protocol yielded superior inter-sample discrimination and stronger in vitro-in vivo correlations, and thus was selected for model development. Under homogenization, AC-based eGI achieved the lowest prediction errors (MAE = 1.25, RMSE 1.50) and perfect rank-order preservation (Spearman's ρ=1.0), followed by AUC₁₂₀ and [Formula: see text] -based eGI. The AC-based eGI was strongly correlated with starch molecular parameters including amylose chain length (p < 0.01), whereas both kinetics-derived predictors lacked significant correlations. Cross-experiment comparison further revealed that identical varieties produced markedly different digestion curves, highlighting the inherent variability of kinetics-based predictors. These findings demonstrate that AC provides a robust and transferable basis for the rapid screening of cooked rice GI.