There is a trend to start dialysis earlier in patients with chronic renal failure. Studies that suggest improved survival from earlier initiation of dialysis are flawed in that they have measured survival from start of dialysis rather than from a time point before dialysis, when patients have the same renal function. This flaw is termed lead-time bias. Using the electronic patient record at the renal unit of Glasgow Royal Infirmary, all patients were identified who had received dialysis for chronic renal failure and who had sufficient data to calculate the time point at which they reached an estimated creatinine clearance (eC(Cr)) of 20 ml/min (n = 275). This date was used to time survival. The patients were divided into early and late start groups by the median eC(Cr) for all patients at initiation of dialysis, which was 8.3 ml/min. There was no significant benefit in patient survival from earlier initiation of dialysis. A Cox proportional hazards model demonstrated a significant inverse relationship between eC(Cr) at start of dialysis and survival (hazard ratio, 1.1; P = 0.02), i.e., patients who started dialysis with a lower eC(Cr) tended to survive longer. This relationship retained significance when gender, age, weight, presence of diabetes, mode of first dialysis, initial dialysis access, hemoglobin, serum albumin, blood leukocyte count, Wright/Khan index, and eC(Cr) at the start of dialysis were taken into account. This study fails to support a policy of earlier initiation of dialysis for patients with end-stage renal failure.