Comparison of blood pressure machines

After examining published papers and our own experiences we suggest some simple guidelines for future studies which compare sphygmomanometers. The sample should be stratified to cover a wide range of blood pressures. Observers should measure pressures using a balanced cross-over design under standard conditions. Data should be shown on a scatterplot to allow visual estimation of bias, variability of readings and non-linear relationships. Statistical analyses if universally applied would allow comparison of studies. We suggest the use of the following common methods; the mean of the differences between paired readings with standard deviation, linear regression with both zero and non-zero intercepts and analysis of variance.