As under-reporting of dietary intake, particularly by overweight and obese subjects, is common in dietary surveys, biases inherent in the use of self-reported dietary information may distort true diet-obesity relationships or even create spurious ones. However, empirical evidence of this possibility is limited. The present cross-sectional study compared the relationships of 24 h urine-derived and self-reported intakes of Na, K and protein with obesity. A total of 1043 Japanese women aged 18-22 years completed a 24 h urine collection and a self-administered diet history questionnaire. After adjustment for potential confounders, 24 h urine-derived Na intake was associated with a higher risk of general obesity (BMI≥25 kg/m2) and abdominal obesity (waist circumference≥80 cm; both P for trend=0·04). For 24 h urine-derived protein intake, positive associations with general and abdominal obesity were observed (P for trend=0·02 and 0·053, respectively). For 24 h urine-derived K intake, there was an inverse association with abdominal obesity (P for trend=0·01). Conversely, when self-reported dietary information was used, only inverse associations between K intake and general and abdominal obesity were observed (P for trend=0·04 and 0·02, respectively), with no associations of Na or protein intake. In conclusion, we found positive associations of Na and protein intakes and inverse associations of K intake with obesity when using 24 h urinary excretion for estimating dietary intakes. However, no association was observed based on using self-reported dietary intakes, except for inverse association of K intake, suggesting that the ability of self-reported dietary information using the diet history questionnaire for investigating diet-obesity relationships is limited.