The agreement between the Reflotron dry chemistry analyser and laboratory methods for measurement of total cholesterol and triglycerides was assessed by studying the between-method variability for individual samples. A paired comparison of replicate measurements from venous blood was made for 105 cholesterol and 90 triglyceride specimens. There was no significant mean difference between the two methods for cholesterol measurement (difference 0.02 mmol l-1, 95% Cl -0.07 to 0.10 mmol l-1) but there were wide limits of agreement (+/- 2SD) of -0.87 to 0.90 mmol l-1. Some of the differences in measurement were large enough to be clinically misleading and would result in inconsistencies in diagnostic assignment between the two methods. Both methods achieved relatively poor repeatability which limited the agreement possible (coefficient of repeatability, defined as 2SD of the differences, 0.59 mmol l-1 for the Reflotron and 0.38 mmol l-1 for the laboratory). The repeatability of triglyceride measurement was similar for the Reflotron and laboratory (coefficient of repeatability 0.24 mmol l-1 and 0.27 mmol l-1, respectively). However, there was a significant systematic difference between methods, with the Reflotron reading 0.30 mmol l-1 less than the laboratory method (95% Cl 0.26 to 0.35 mmol l-1) and the limits of agreement between the methods were -0.09 to 0.70 mmol l-1. It is important that clinicians recognize the limits of agreement between dry chemistry analysers and laboratory methods and take these into account when using them to screen for hyperlipidaemia or to monitor treatment of hyperlipidaemic diabetic patients.