How is biological variation used to set analytical performance specifications?

Understand what is varying

Even when a person is in a stable state, their measurand can vary from one sample to the next. Within-person biological variation, often expressed as CV_I, describes that variation relative to the person’s level. Between-person biological variation describes differences between people. Analytical imprecision contributes a separate source of variation to the observed laboratory results. [1] [2]

The reasoning behind a biological variation–based imprecision target is that analytical scatter should be small relative to the within-person variation it accompanies. That can be useful for measurands where biological variation is a sensible basis for the use being considered. An estimate taken from a different specimen type, population or sampling interval may describe a different situation. [1] [2]

Turn one estimate into an imprecision target

Suppose a relevant and credible within-person CV estimate for a fictional measurand is 6.0%. The conventional desirable imprecision target is analytical CV no greater than half CV_I. In this example, 0.5 × 6.0% gives a target of 3.0%. The factor is part of the specification model. It does not come from the laboratory’s own precision study. [2] [3]

Now compare the target with a within-laboratory analytical CV estimate of 2.5% at a relevant concentration. The point estimate is below the 3.0% target. A laboratory assessing a real method would check whether the analytical study represents the procedure and intended conditions, and how precisely the CV was estimated.

Check the evidence before adopting the target

Look up the measurand in a curated source such as the EFLM Biological Variation Database. Check the specimen, population, sampling interval and quality of the evidence behind the estimate. Ask whether the clinical use relies on the kind of variation the estimate describes. Biological variation estimates also have uncertainty, so note the source and what would make you revisit the choice. [1] [2]

Keep imprecision and bias separate

This calculation sets an imprecision target. Biological variation can also be used to derive a specification for bias, but that addresses systematic difference and uses different inputs in conventional models. State which performance characteristic the target applies to. For a decision tied to a particular clinical threshold, consider whether an outcome-based approach would be more informative. [1] [2]

Example wordingUsing an illustrative within-person CV of 6.0% and the conventional desirable factor of 0.5 gives an analytical imprecision target of 3.0% CV. The observed within-laboratory CV of 2.5% is below this target at the concentration and under the conditions studied.