AIMS/INTRODUCTION: We carried out a cross-sectional study of people with type 2 diabetes mellitus to elucidate the association between sleep duration and food intake. MATERIALS AND METHODS: Overall, 2,887 participants with type 2 diabetes mellitus (mean age 63.0 years; 61.1% men; mean glycated hemoglobin level 7.5%) were included in this study. The participants' self-reported dietary habits and sleep duration were evaluated using a brief self-administered dietary history questionnaire and Pittsburgh Sleep Quality Index, respectively. The participants were categorized into the following four groups based on sleep duration: <6, 6-6.9, 7-7.9 (reference) and ≥8 h. RESULTS: No significant differences were observed between the groups regarding energy intake (kcal/day), absolute intake (g/day) or relative intake (% energy) of carbohydrates, total fat, proteins and fibers. However, confectionery intake was higher in the <6 h group and lower in the ≥8 h group than in the reference group after adjustment for confounding factors. In multivariate analysis, sleep durations <6 h and ≥8 h significantly correlated with increased (95% confidence interval 0.55 to 3.6; P = 0.0078) and decreased (95% confidence interval -4.0 to -0.32; P = 0.021) confectionery intake, respectively. Confectionery intake was positively correlated with female sex, glycated hemoglobin level and dyslipidemia, whereas it was negatively correlated with alcohol consumption and current smoking status. CONCLUSIONS: Short sleep duration is associated with high confectionery intake in people with type 2 diabetes mellitus; this might disturb their glycemic control. Therefore, short sleepers with type 2 diabetes mellitus could improve their glycemic control by avoiding confectionery intake and maintaining adequate sleep duration.