Interest in the dietary glycaemic index (GI) and glycaemic load (GL) as risk factors for chronic diseases has grown in recent years but findings have been controversial. We describe the compilation of the GI database for the cohort studies within the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study and the main characteristics associating with diet GI and GL. The ATBC Study enrolled 29 133 male smokers aged 50–69 years who filled in a dietary history questionnaire upon study entry. The dietary data included 1097 foods, of which 195 foods with no or a negligible amount of available carbohydrates were assigned a GI of zero. Based on preset methodological criteria for published GI studies, the GI value of a similar food was available for 130 foods, and the GI of related food was assigned to 360 foods. The GI values of these foods served in the GI calculation of 412 composite foods. The median diet GI among the ATBC Study participants was 67·3 (interquartile range 64·8–70·0), and the median diet GL was 175 (interquartile range 158–192). The intakes of carbohydrates, protein and fat decreased, and the intake of fibre increased, with increasing GI. The GL showed a positive correlation with intakes of carbohydrates and dietary fibre and a negative correlation with intakes of protein and fat. The GI studies available that fulfilled the minimum methodological requirements cover a sufficient amount of foods to form a meaningful GI database for epidemiological study. This, however, requires the availability of GI values for relevant local carbohydrate-containing foods.