Most studies analyzing diet-disease relations focus on single dietary factors rather than combining different nutrients into the same statistical model. The objective of this study was to identify dietary factors associated with the probability of having diabetes identified by screening (SDM) in Danish men and women aged 30-60 y. A specific objective was to examine whether an alternative statistical approach could provide additional information to already existing statistical approaches used in nutritional epidemiology. Baseline data from the Danish population-based Inter99 study were used. The dietary intake of 262 individuals with SDM was compared with that of 4627 individuals with normal glucose tolerance (NGT) using 2 different types of multiple logistic regression models adjusted for potential confounders. The first model included single dietary factors, whereas the second model was based on substitution of macronutrients. In the models with single dietary factors, high intakes of carbohydrates, dietary fiber, and coffee were inversely associated with SDM (P < 0.01), whereas high intakes of total fat and saturated fat were positively associated with SDM (P < 0.05). A modest U-shaped association was found between alcohol consumption and SDM (P = 0.10) [corrected] Results from the substitution model showed that when 3% of energy (En%) as carbohydrate replaced 3 En% fat or alcohol, the probability of having SDM decreased by 9 and 10%, respectively (P < 0.01) [corrected] No other macronutrient substitutions resulted in significant associations. Hence, the statistical approach based on substitution of macronutrients provided additional information to the model analyzing single dietary factors.