Background: Dietary carbohydrates are the major energy source in the Mediterranean region. While the quality of carbohydrate-containing foods has been measured using the glycemic index (GI), a representative database of dietary glycemic index (dGI) and load (dGL) that incorporates both the quantity and quality of carbohydrates in that region is lacking. Objectives: To develop a harmonized database of dGI and dGL for adults who participated in national surveys, and to assess the validity of different methods in determining GI values for foods by content experts and artificial intelligence (AI). Methods: Food consumption data were obtained from nationally representative surveys in Croatia, France, Italy, Montenegro, Portugal, and Spain. Three independent approaches were used to assign GI values to carbohydrate-containing foods: manual assignment by a nutritionist, an AI-enabled approach based on the textual similarity of food descriptions, and an AI-assisted expert-curation assignment. For each country, dGI and dGL were estimated based on a total of 956 carbohydrate-containing foods, from which the top dGL-contributing foods were identified. The triad method was used to calculate the correlations between each of the three independent methods and the underlying unknown. Results: Dietary data of 14,020 participants from seven surveys (all used two 24-hour recalls except Italy [3-day diet record]) were analyzed. Mean population age ranged from 43.4 to 57.5 years; mean body mass index from 24.5 to 27.2 kg/m²; and mean carbohydrate intake from 182.0 to 256.1 g/day. dGI ranged from 52.1 (95%CI: 51.7-52.4) in Spain to 57.8 (95%CI:57.7-58.0) in Italy 2005-2006, energy-adjusted dGL ranged from 103.3 (95%CI:101.9-104.6) in Croatia to 149.3 (95%CI:148.4-150.3) in Italy 2005-2006. Men had a higher mean dGI 57.5 (95%CI:57.4-57.6) compared with women 55.5 (95%CI:55.4-55.6), whereas women had a higher energy-adjusted dGL 127.9 (95%CI:127.4-128.5) than men 125.6 (95%CI:124.8-126.4). Across surveys, the top 3 dGL-contributing foods were wheat bread, potatoes, and pasta. Conclusions: This study developed the first harmonized database of dGI and dGL based on nationally representative dietary surveys and conducted the first validation of AI-based GI assignment in a multi-country Mediterranean setting. The database could serve as a standardized resource for nutritional surveillance and epidemiological research on carbohydrate quality and diet-related health outcomes in Mediterranean populations.
Development of a representative database for dietary glycemic load across countries in the Mediterranean region / A. D’Angelo, B. Yang, K. Della Corte, D. Della Corte, C.V.B. La Vecchia, L.S.A. Augustin, S. Liu. 43. International Symposium on Diabetes & Nutrition Tarragona, Spain 2026.
Development of a representative database for dietary glycemic load across countries in the Mediterranean region
A. D’AngeloPrimo
;C.V.B. La Vecchia;
2026
Abstract
Background: Dietary carbohydrates are the major energy source in the Mediterranean region. While the quality of carbohydrate-containing foods has been measured using the glycemic index (GI), a representative database of dietary glycemic index (dGI) and load (dGL) that incorporates both the quantity and quality of carbohydrates in that region is lacking. Objectives: To develop a harmonized database of dGI and dGL for adults who participated in national surveys, and to assess the validity of different methods in determining GI values for foods by content experts and artificial intelligence (AI). Methods: Food consumption data were obtained from nationally representative surveys in Croatia, France, Italy, Montenegro, Portugal, and Spain. Three independent approaches were used to assign GI values to carbohydrate-containing foods: manual assignment by a nutritionist, an AI-enabled approach based on the textual similarity of food descriptions, and an AI-assisted expert-curation assignment. For each country, dGI and dGL were estimated based on a total of 956 carbohydrate-containing foods, from which the top dGL-contributing foods were identified. The triad method was used to calculate the correlations between each of the three independent methods and the underlying unknown. Results: Dietary data of 14,020 participants from seven surveys (all used two 24-hour recalls except Italy [3-day diet record]) were analyzed. Mean population age ranged from 43.4 to 57.5 years; mean body mass index from 24.5 to 27.2 kg/m²; and mean carbohydrate intake from 182.0 to 256.1 g/day. dGI ranged from 52.1 (95%CI: 51.7-52.4) in Spain to 57.8 (95%CI:57.7-58.0) in Italy 2005-2006, energy-adjusted dGL ranged from 103.3 (95%CI:101.9-104.6) in Croatia to 149.3 (95%CI:148.4-150.3) in Italy 2005-2006. Men had a higher mean dGI 57.5 (95%CI:57.4-57.6) compared with women 55.5 (95%CI:55.4-55.6), whereas women had a higher energy-adjusted dGL 127.9 (95%CI:127.4-128.5) than men 125.6 (95%CI:124.8-126.4). Across surveys, the top 3 dGL-contributing foods were wheat bread, potatoes, and pasta. Conclusions: This study developed the first harmonized database of dGI and dGL based on nationally representative dietary surveys and conducted the first validation of AI-based GI assignment in a multi-country Mediterranean setting. The database could serve as a standardized resource for nutritional surveillance and epidemiological research on carbohydrate quality and diet-related health outcomes in Mediterranean populations.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.




