OBJECTIVE We tested the hypothesis that there is no difference in the change in serum creatinine level following computed tomography (CT) between those given high osmolality contrast, low osmolality contrast and those not receiving contrast material. METHODS Patients were assigned according to current radiological practice to receive one or other type of radiocontrast material or to have a scan without intravenous contrast (plain scan). METHODS The CT unit of Royal Newcastle Hospital, a tertiary referral institution. METHODS Of 3188 inpatients having a CT scan between June 1988 and December 1989, 1041 patients were eligible (having a first scan in "office hours" outside holiday periods, not due for imminent discharge, aged 18 years or more and not presenting to the Intensive Care Unit or with acute trauma). Twenty-five patients were excluded as baseline measures of renal function were missing and 132 subjects were lost to follow-up, leaving 884 study subjects. METHODS CT scans using high osmolality, low osmolality or no contrast (plain scan). METHODS Renal impairment as defined by a maximal increase in the serum creatinine level of greater than or equal to 50% or greater than 0.04 mmol/L from the baseline level on at least one of the subsequent four days. RESULTS Renal impairment was seen in 4% (12 of 292), 12% (23 of 187) and 4% (16 of 405) of patients given high osmolality, low osmolality or no contrast respectively. Age and the baseline level of serum creatinine were independent predictors of the development of renal impairment (P = 0.04 and 0.02 respectively) and those given low osmolality contrast were 3.2 times (95% confidence interval, 1.6-6.3) more likely to develop renal impairment than those given no contrast. There was no excess risk with the use of high osmolality contrast compared to no contrast (odds ratio, 1.06; 95% confidence interval, 0.5-2.3). Selection factors (sicker patients being given low osmolality contrast) are likely to have accounted for the excess risk in the low osmolality group as mortality in hospital was higher in this group than in the others. The 51 patients who developed renal impairment (cases) were matched for age, sex, type of contrast and pre-existing renal impairment with up to three controls (150 patients in total). Cases were more likely to have had a blood transfusion (odds ratio, 6.40; 95% confidence interval, 2.18-22.63) or surgery (odds ratio, 3.22; 95% confidence interval, 1.19-7.65) than controls. CONCLUSIONS Confounding by other factors which impair renal function is likely to explain previous suggestions of an effect of radio-contrast material on renal function. There does not appear to be a risk of renal impairment from the use of high osmolality radiocontrast material (although a small effect or an effect in particular subgroups cannot be excluded by our study). Fear of causing or exacerbating renal damage should not be a reason to use low osmolality contrast material, nor should it be a reason for with-holding contrast studies.