The increasing recognition of the adverse consequences of anaemia and the observation that anaemia starts fairly early during the progression of chronic kidney disease has led to an increasing interest in anaemia correction of patients not (yet) on dialysis. One of the important questions in this context is, whether anaemia and / or its correction have an impact on the decline in renal function. The evidence that has accumulated over many years seems to allow three conclusions. First, anaemia and /or its correction do not have a "major" impact on the progression of renal disease. Second, an effect in either direction, that is not obvious on the basis of available data, but may still be clinically relevant, can not be excluded. Third, providing a definite answer to this question is difficult. A recent study by Becker et al. in transplant recipients (NDT 17, 2002: 1667-1673) confirms all three of these conclusions. The authors have retrospectively analysed a transplant database and identified patients that were started on rHuEPO therapy at different time intervals after transplantation. In a group of patients receiving rHuEPO not before almost 10 months after transplantation they found a decrease in the slope of serum creatinine. Moreover, they observed, that the graft survival in these patients tended to be lower than in a group of "control" patients not receiving anaemia therapy. This difference, however, did not reach the level of significance. The authors frankly discuss and acknowledge the methodological flaws and limitations of such a study design. One can only underline their conclusion that prospective trials are needed to provide further information. Yet, it is important to also recognize the difficulties of such an approach. Given the proven benefits of anaemia correction, it appears unjustifiable to leave a group of patients untreated in a prospective trial and compare them with others, in whom anaemia is corrected, even though in reality anaemia is frequently not addressed. Thus it is only possible to compare different target levels, and the smaller the difference in achievable haemoglobin levels, the larger the patient number that has to be studied to detect differences.