Introduction Epidemiologic evidence on the role of dietary intake of carbohydrates in development of type-2-diabetes is inconclusive.

Methods We conducted a systematic review of studies reporting the association between dietary intake of carbohydrate and its subtypes (starch, sucrose, glucose, fructose, lactose, and maltose) and risk of incident diabetes. We searched MEDLINE (1966 to October 2010) and hand searched bibliographies of retrieved articles. Studies were included if they had a prospective design, adult population, assessed dietary intake at baseline, and recruited participants free of diabetes at entry. Risk estimates (RR, HR, OR) from models with the highest degree of multivariate adjustment in each study were transformed to a standardised top-vs-bottom fifth estimate according to the population's baseline distribution of each nutrient's values. We used the I² statistic to measure heterogeneity between studies and calculated pooled risk estimates for incident diabetes with random-effects meta-analysis.

Results Ten prospective cohort studies with data on 420 840 participants and 11 517 incident diabetes events were included. Highest to lowest fifth of intake of sucrose was associated with a 15% lower risk of diabetes (RR: 0.85, 95% CI 0.75 to 0.97). Other carbohydrate subtypes were not significantly associated with diabetes risk.

Conclusion All studies reported risk estimates adjusted for total energy intake and thus model an iso-energetic diet. Lower risk of diabetes associated with higher intake of sucrose is most likely to reflect the effect of substitution of sucrose for other nutrients rather than net increased intake of sucrose itself. Nutrient substitution patterns require further investigation.