The incidence of diabetes in Indonesia has perpetually increased at significant rate, from 1.4% in 2013 to 4.9% in 2018. Consumption of high glycemic index (GI) foods closely related to the rise of this case. On the other hand, numerous studies focusing on GI of various foods and their chemical characteristics of carbohydratebased foods such as resistant starch, fiber, fat, protein, phenol, and flavonoid content have remained indecisive conclusion regarding to correlation between consumption of these diets and lower risk of diabetes. In this case, meta-analysis is obviously required for producing novel insights into preparation of controlled GI. This present work aimed at generating the food model that has controlled GI using verified data reproduced from meta-analysis approach. The research involved a systematic review and meta-analysis approach, carried out in 5 stages. First, starchy-based foods with low GI were selected. Second, two major chemical characteristics contributing to GI of the foods were determined. Third, the source of main determinants of the chemical characteristics was decided. Fourth, GI-lowering mechanisms of the determinants were determined. Fifth, association between the chemical determinants and GI values was verified using in vivo experiment. The first four stages were accomplished by collecting secondary data published in various journals, conducting systematic reviews, assessing the quality of selected journals by GRADE (Grading of Recommendations Assessment, Development, and Evaluation) method, and performing meta-analysis. For the ultimate step, food modelling was constructed according to meta-analysis and in vivo experiment results. The data was sourced from 500 scientific literatures, consisting of accredited national journals (16 titles) and internationally reputed journals (484 titles) collected from journal databases including Proquest, Science Direct, Ebsco, Cengage Library, and Google Scholar up to July 2019. Tittles and abstracts of these articles were screened using Zotero and Mendeley software. From systematic review, 57 selected articles were obtained for meta-analysis. In meta-analysis, calculation of key statistical values were performed including Hedge's d effect size value, 95% confidence interval (CI), weight of each study (% weight), I2 value, z value, and p value using Microsoft Excel 2019. These values were then plotted into forest and funnel plot using Microsoft Word 2019 and metaessential tool software, respectively. As the results, meta-analysis revealed that the food crop having the lowest GI was attributed to legumes. Furthermore, within legume subgroup analysis, species that showed significant low GI was sorghum (SMD / standard mean difference: - 0.69, 95% CI: −2.33 to 0.96, p <0.001), and red kidney bean (SMD: −0.39, 95% CI: −2.37 to 1.59, p = 0.001). The two significant factors responsible for the GI of starchy foods included presence of resistant starch (RS) and phenolic content (respectively SMD: −0.72, 95% CI: −1.00 to .40.43, p <0.001; SMD: - 0.67, 95% CI: −1.05 to −0.30, p <0.001). In addition, the main source for the chemical determinant factors was Pomelo Peel Extract/ PPE (SMD: −3.14, 95% CI: −5.53 to -0.75, p <0.001). Meanwhile, the mechanism of GI-lowering effect was based on the -amylase inhibition (SMD: −0.87, 95% CI: −1.29 to -0.45, p <0.001). Inhibition of the -amylase enzyme as a result of this study was more effective in reducing IG by 4.5 times compared to inhibition of the -glucosidase enzyme. The result of meta-analysis was verified using in vivo test according to protocol of ISO 2664: 2010. The experiment involved 10 healthy subjects, while GI value was calculated according to the percentage of area under the glycemic response curve of the tested food compared to glucose as reference (carried out duplicates in different days with CV ≤30%). Blood glucose level was checked using glucometer with the finger prick test performed at following intervals: 0, 15, 30, 45, 60, 90, and 120 minutes after meals. The food used for model in the in vivo verification was comprised of 5 groups: (1) non-treated carbohydrate-based foods, (2) cooked rice “pandan wangi” added with resistant starch (RS), (3) cooked rice “pandan wangi” added with phenol, (4) cooked rice “pandan wangi” added with RS and phenol, and (5) red beans with the addition of RS and phenol. For group 1, the food model included cooked rice “pandan wangi” (high GI), red sorghum (medium GI), and red beans (low GI). Group 2 included cooked rice “pandan wangi” enriched with RS (2.3%). For group 3, cooked rice “pandan wangi” was added with either Pomelo peel extract – PPE (274 mg GAE/100g) or bay leaf extract – BLE (556 mg/100g). For group 4, the rice was treated with addition of RS (2.3%) and PPE (274 mg GAE/100 g) and addition of RS (2.3%) and BLE (556 mg/100g). The last group, red bean was added with RS (2.3%) and PPE (274 mg GAE/100g). The results demonstrated that carbohydrate-based foods in group 1 consistently showed three levels (high, medium, low) of GI, i.e. rice “pandan wangi” (91), steamed red sorghum (66), and steamed red beans (20), respectively. Regarding to group 2 and 3, the supplementation of RS to high GI foods less effective to attenuate GI value, but the addition of phenolic compound to the food models effectively reduced GI value up to 20-26%. For group 4 and 5, the addition of RS and phenolic compound to high GI foods resulted in a meaningful reduction of GI, reaching up to 24-26%, but the effect was not found in low GI foods. Ultimately, the results of in vivo experiment clearly denoted the opportunity of reducing GI in high GI starchy foods through enrichment of RS and phenolic compounds. As a conclusive remark, meta-analysis constitutes a reliable approach for selecting, classifying, and facilitating the decision making process in determining the treatments for in vivo food model verification.

Keywords: food model, glycemic index, in vivo, meta-analysis, systematic review.