Figure 1

Systematic search for eligible studies of dietary sugar intake and type 2 diabetes. (a) 2,005 from PubMed and 1,143 from Scopus data base searches. (b) Titles remotely on topic were screened twice. (c) We completed a full-text review of all studies of dietary patterns, glycemic load/index, and carbohydrates to assess whether a measure of dietary sugar was examined individually. We also reviewed the full text and bibliographies of studies of sugar-sweetened beverages (SSB), juices, sugars, macronutrients and key reviews and commentary. (d) We identified three cohorts with multiple publications, from which we selected for this synthesis the one publication in which SSB was either the main study variable or the definition was the clearest. We identified two publications of the Health Professionals Follow-up study (HPFS); of these two publications, the one that assessed SSB as the primary study variable was selected for inclusion [34] and the other that presented analyses stratified by the main variable, caffeine consumption, was excluded [35]. We selected one of the three publications from the Nurse’s Health Study (NHS). Bazzano and coworkers [39] reported risk separately for a one-increment serving of sugar-sweetened colas, fruit punch, low calorie cola, and other carbonated beverage. In a personal communication from a 2010 meta-analysis [50], Malik and coworkers report a risk estimate for SSB intake, but the definition was not provided nor was the analysis adjusted for age. Although not ideal, the Bhupathiraju et al. analysis of SSB, stratified by caffeinated and caffeine-free beverage consumption, provides a clear definition (sugar-sweetened carbonated beverages) and analysis, and therefore was selected for inclusion in this paper [35]. Our final exclusion was a 2013 publication of EPIC-France [31], from which all participants were represented by an included EPIC publication [29].