Figure 2.

Process of the review and outcomes of the inclusion/exclusion criteria for the T2D–GI risk relation. The literature was searched for prospective cohort studies investigating this relation, from 1946 to 6 December 2018. MEDLINE, EMBASE, and other sources (see Methods) were searched. Superscripts: a. Whether T2D is related to the dietary glycemic index (GI) and whether the risk relation is >1.20 with a lower confidence limit > 1.10 for each 10th to the 90th percentile of GI intake (which on analysis was per 10 units GI). b. Is the T2D–GI risk relation conditional on the validity of the dietary instrument, on the sex of the participants, or on the BMI (kg/m²)? c. Details of the full reports excluded are described in the associated text and in more detail in Section 3 (f and g) of the Supplementary Materials. d. Studies with dietary instrument correlations for carbohydrate ≤ 0.55 were invalid. e. When a study was suspected of being an outlier was proven to be so upon analysis (p < 0.05) and persisted to be so in several different meta-analyses (Table S15), it was excluded from the primary analysis. Abbreviations: DRM, dose-response meta-analysis; GI, glycemic index; n, number of published reports; k, number of studies (more than one may appear in one report); T2D, type 2 diabetes.