The aim of this 2-year prospective cohort study was to determine whether food-level sugar-starch combinations are predictors of dental caries in a low-risk adolescent population. A total of 645 subjects, aged 12-13 yr at baseline, were recruited from 25 secondary colleges in the north-west region of metropolitan Melbourne, Australia. Examinations to record dental caries status were conducted annually. Dental caries was diagnosed according to the criteria of the World Health Organization. Dietary information was collected by four, continuous 4-d records. Demographic data was collected by parental self-administered questionnaire. A total of 504 subjects provided complete information for analysis. Approximately 37% of subjects experienced an increment in caries. In the multivariate model, only the low sugar-high starch food group was a significant predictor of caries increment on all surfaces and pit and fissure surfaces. For both these surfaces, significant interactions with starch at low sugar and across those clusters with a maximum proportion of sugar and/or starch (that is, high sugar-low starch, medium sugar-medium starch and high sugar-low starch) were found. Sugar-starch interactions may be predictive of caries risk in a low-risk adolescent population. Changing patterns of food consumption and the widespread exposure to various fluoride vehicles are possibly altering the diet-dental caries dynamic that once existed.