In vitro studies have suggested that a low packed cell volume may cause reagent strips to overestimate blood glucose concentration. The reliability of such strips has therefore been investigated in diabetic patients with end-stage renal failure, all of whom had low haematocrits (24 +2- 4 (4/- SD) %). Three brands of strips were evaluated and their results were compared with laboratory whole blood glucose concentrations determined on the same samples. All over-read compared with the laboratory (BM Glycemie 1-44 strips 118 +/- 1%, Glucostix 132 +/- 2%, ExacTech strips 163 +/- 4%) and the percentage overread was negatively correlated with the haematocrit for each strip (all p less than 0.001).