Test result variation and the quality of evidence-based clinical guidelines

There is a plethora of supposedly evidence-based published clinical guidelines, most often prepared under the auspices of professional bodies. Many guidelines contain numerical laboratory test results as criteria for clinical action, very often simply quoted as single numbers. Every test result is subject to a number of sources of variation. Analytical imprecision and within-subject biological variation are particularly important. The influence of both of these on the dispersion of a single test result and on the number of samples required to make clinical decisions can be easily calculated using simple formulae. The effect of performing replicate analyses of one sample and of taking multiple samples can also be easily investigated. Authors of scientific statements, clinical guidelines, and practice recommendations should undertake such calculations before promulgating their efforts in the public domain. Guidelines on cholesterol and high sensitivity C-reactive protein have been examined using these approaches and these investigations allow the following conclusions.

Analytical imprecision should be made low. If analytical imprecision is generally greater than biological variation, then reduction in analytical imprecision is valuable. If biological variation is greater than imprecision, then collection of more than one sample from an individual prior to decision-making is useful.