Measurement Procedure Comparison in the Clinical Lab

Measurement procedure comparison is one of the most common techniques used by both manufacturers and medical laboratorians to estimate the bias of an in vitro diagnostic (IVD) measurement procedure relative to a comparator. It involves the comparison of results from patient samples from two measurement procedures intended to measure the same component (eg, measurand concentration) with the key determination being the estimate of bias between them.

Several different scenarios exist in which measurement procedure comparison studies are indicated. For both the manufacturer and the medical laboratorian, the ideal scenario is the comparison of a candidate measurement procedure to a generally accepted standard or reference measurement procedure.

The measurement procedure comparison experiment gives an estimate of the bias between two measurement procedures and estimates for bias at any specified concentration. The estimates of the concentration for a sample from the comparative measurement procedure should ideally have the following characteristics:

• Have lower uncertainty than estimates of the concentration for that same sample from the candidate measurement procedure, which can often be achieved by averaging replicates, when needed.
• Be free from known interferences, whenever possible. This should be true for both the comparative and the candidate measurement procedures.
• Use the same units as the candidate measurement procedure or are able to be converted to the same units.
• Be traceable to reference material or reference measurement procedures, whenever possible.

Clinical and Laboratory Standards Institute guideline EP09—Measurement Procedure Comparison and Bias Estimation Using Patient Samples is written for laboratorians and manufacturers and describes procedures for determining the bias between two measurement procedures, and it identifies factors for consideration when designing and analyzing a measurement procedure comparison experiment using patient samples.