One hundred and eighty-seven diabetic and 105 control subjects collected timed overnight urine samples to measure the inter-individual variation in creatinine excretion rate and its determinants, and to test the relationship between albumin excretion rate (AER) and two 'surrogate measures', the albumin concentration and albumin:creatinine ratio. Creatinine excretion was 55% higher in men than women (geometric mean 8.9 mumol min-1 (95% confidence limits 4.7-17.0) compared with 5.7 (3.0-10.9); p < 0.001). Gender accounted for 31% of the variation in creatinine excretion and body mass index 1.4%; neither age nor the diabetic state had a significant effect. The relationships between AER and the two surrogate measures differed between diabetic subjects and controls such that relationships constructed from non-diabetic data would not hold true for diabetes. Likewise, the relationship between AER and albumin:creatinine ratio differed between men and women such that a ratio of 4.0 mg mmol-1 corresponded to a predicted AER of 35 micrograms min-1 in men and 23 micrograms min-1 in women. The albumin:creatinine ratio outperformed albumin concentration in terms of sensitivity and specificity and its performance was better in women than men. We conclude that the albumin:creatinine ratio is a better surrogate for AER than albumin concentration. If 'action levels' are to be defined for screening programmes, they should be derived from diabetic and not non-diabetic data and should be different in men and women. We propose a direct rather than screening role for the albumin:creatinine ratio in the management of diabetic nephropathy.