Renal failure is increasingly being recognised in CF patients, usually as a consequence of long-term nephrotoxic therapy. There is a need for a simple method of assessment of renal function in this patient group. We compared measured creatinine clearance from validated timed urine collections (the generally accepted practical test of glomerular filtration) with 10 formulae used to estimate creatinine clearance in a group of 74 CF adult patients and 29 matched normal controls. Compared to direct measurement, formulae gave a range of values (95% CI for mean bias -13 to +27.9 ml/min). Even those with the best correlation (r=0.7) gave wide error ranges (limits of agreement: -42.3 to 45.9 ml/min). The most commonly used formulae (Cockroft-Gault [CGF] and abbreviated Modification of Diet in Renal Disease [aMDRD]) were not superior to most other formulae tested. Both CGF and aMDRD-derived estimates compared less favourably in CF patients than controls (mean bias: 9.7 vs 3.4 ml/min (p<0.05) and 4.9 vs 1.4 (p<0.05) respectively; 78% vs 95% (p<0.01) and 77% vs 97% (p<0.01) of estimates within 33% of measurement respectively). In particular, both CGF and aMDRD grossly overestimated renal function (mean bias 18.3 and 15.8 ml/min respectively, p<0.001) in CF patients with reduced creatinine clearance (<80 ml/min). CGF, aMDRD and other formulae cannot be used to reliably assess renal function in CF patients, since they will fail to detect those with renal impairment. Some form of carefully supervised direct measurement is still required.