In patients with type 2 diabetes mellitus (DM), the factors associated with under- or overreporting of protein intake in nutrition assessment tools, as well as the variability of diet records, have not been fully established. The aim of this cross-sectional study was to evaluate factors associated with under- or overreporting of protein intake and its variability in patients with type 2 DM. Protein intake was estimated in 205 patients (aged 59.8+/-9.6 years) using 3-day weighed diet records and 24-hour nitrogen output (criterion standard). Twenty-three patients repeated the 3-day weighed diet records three times. Clinical, nutrition, and lifestyle evaluations were performed. Coefficients of variation were calculated for protein intake. Factors associated with under- and overreporting were assessed using multivariate logistic regression models. Coefficients of variation for protein intake estimated by weighed diet records or nitrogen output were similar (11.9% vs 11.3%; P>0.05). Using Beaton's formula, a difference of 16.5% in protein intake between two 3-day weighed diet records was acceptable. The lowest A1c test tertile (< or =6.9%) was associated with protein intake underreporting (odds ratio [OR]=0.40; 95% confidence interval [CI]=0.16 to 0.99; P=0.046] after adjustment for sex, age, employment status, and living alone. Male sex (OR=6.66; 95% CI: 2.08 to 22.07; P=0.002), A1c test (OR=1.29; 95% CI: 1.02 to 1.64; P=0.036), and body mass index (OR=0.89; 95% CI: 0.80 to 0.994; P=0.039), adjusted for physical and employment status, education, and preparing one's own meals, were associated with overreporting. In conclusion, in patients with type 2 DM, a difference >16.5% in protein intake between two 3-day weighed diet records should be interpreted as a true discrepancy. Poor glucose control and male sex increase the chance of inaccurate 3-day weighed diet records.