A method of analysis of laboratory data in an epidemiological study where time trends are present

Abstract

Where data are collected over an extended period, in either cross‐sectional or prospective studies, time trends in laboratory measurements can have misleading effects on observed associations, especially where it has not been possible to randomize the order of entry of participants to the study. In an ongoing prospective study of ischaemic heart disease, a modified cusum technique is used to examine the data for time trends. The paper considers a method for estimating where changes may have occurred, a method for adjusting for such changes in investigating the associations between variables and the appropriateness of making such adjustments.