The end of the p value?

The application of statistical methods to medical data has been undergoing a sea-change.This is of par- ticular importance in cardiology because the current methods that statisticians recommend express the results of studies in terms that are directly relevant to the clinical use to which they may be put.In March 1986 the British Medical Journal nailed its colours firmly to the mast, telling readers that "authors ... will be expected to calculate confidence intervals whenever the data warrant this approach"'2 and the Lancet,"4 Annals of Internal Medicine, and American Journal of Public Health are among other joumals that have endorsed the new orthodoxy.We expect that studies reported in the British Heart Journal will increasingly reflect this approach.The nuts and bolts of calculating the confidence intervals of various types of data are described in a series of articles in the British Medical Journal,"9 and below we review some aspects of the approach that are particularly relevant to papers published in the British Heart Journal. Towards estimation and away from hypothesis testingThe null hypothesis generally states that there is no relation between the variables under study.For example, when the change in cardiac output before and after intervention is analysed the null hypothesis proposes that the average change is zero.It follows that calculation of the p value, which is based on the null hypothesis, is frequently an inappropriate statis- tical method for summarising the analysis of car- diological data.Many published studies do not seriously consider the possibility that an intervention has no effect.When a test intervention has been used the question usually being asked is "how great is its effect?"rather than "does it have an effect?"