This study examined the validity and clinical implications of the assumption of the Modification of Diet in Renal Disease Study (MDRD) formula that age, gender, race, and BUN account for creatinine production (CP). The relationships of MDRD GFR, CP, and nutrition were examined in 1074 Dialysis Morbidity and Mortality Study Wave II patients with reported measured creatinine clearances at initiation of dialysis. Age, gender, race, BUN, and serum creatinine (Scr) were used to calculate MDRD GFR. The measured 24-h urinary creatinine was used to estimate CP. In linear regression, Scr positively correlated with CP independent of age, gender, race, and BUN. Compared with the highest CP quartile, the lowest CP quartile had lower creatinine clearance (5.8 +/- 2.9 versus 11.3 +/- 3.4 ml/min, P <.01) despite lower Scr (5.8 +/- 2.6 versus 8.6 +/- 3.1 mg%, P <.01). There was an excellent correlation between the reciprocal of Scr and the MDRD GFR (r = 0.90). As a result, the MDRD GFR was higher in the lowest CP quartile (10.9 +/- 4.6 versus 7.6 +/- 2.4 ml/min, P <.01). Malnutrition (48% versus 26%, P <.01) was more common in the lowest CP quartile. Each 5-ml/min increase in MDRD GFR was associated with 21% higher odds of malnutrition (P = 0.046) in a multivariable logistic regression, which was abolished by controlling for CP. The fundamental assumption of the MDRD formula is invalid in patients with advanced renal failure, and the use of this formula in these patients might introduce biases.