Prescription of low osmolar contrast to prevent nephrotoxicity in subjects with pre-existing renal impairment is costly and has not been clearly shown to be effective. We entered 249 subjects with a pre-contrast serum creatinine greater than 120 mumol/liter (1.35 mg/dl) having cardiac catheterization or intravenous contrast into a randomized controlled trial comparing high and low osmolar contrast. The outcome assessed was a rise in serum creatinine repeated 48 to 72 hours after contrast. A further 117 patients entered the non-randomized prospective arm of the study. In the randomized study the serum creatinine rose by at least 25% after contrast in 8 of 117 (6.8%) given high and in 5 of 132 (3.8%) given low osmolar contrast (P greater than 0.05, one-tailed 95% confidence interval for the difference 3 to 7.8%). More severe renal failure (greater than 50% increase in serum creatinine) after contrast was uncommon (3.4% with high and 1.5% with low osmolar contrast). A rise in serum creatinine after contrast was significantly associated with the severity of the pre-contrast renal impairment and the presence of diabetes mellitus, but not with type of contrast. Diabetics with a serum creatinine greater than 200 mumol/liter (2.25 mg/dl) pre-contrast had a highest risk of deterioration in renal function after contrast. We conclude that in patients with pre-existing renal impairment the incidence of contrast nephropathy was not significantly different comparing high osmolar and nonionic contrast. The potential benefit of nonionic contrast in moderate renal impairment is likely to be small, but trials in diabetics with severe renal impairment should be undertaken urgently.