Analytical Validation

#Define the measurement and its scope

Analytical validation concerns the measurement process itself. It asks whether a method measures the intended target with acceptable performance under defined conditions. The target might be a substance, a physical signal or another clearly specified quantity. The evaluation must match the kind of measurement and its intended use.

Define the sample or input, equipment, software version, processing steps and reported result. State the conditions under which the performance claim applies. Evidence from one sample type, device configuration or operating range does not automatically establish performance for another. Acceptance criteria should be justified and set before results are reviewed where possible.

#Examine relevant aspects of performance

Precision describes how closely repeated measurements agree. Bias describes a systematic difference from a suitable reference value. A method can be precise but biased, so repeatability alone is not enough. Evaluations may also examine performance across the measurement range and the smallest amounts that can be detected or measured reliably.

Other relevant questions include whether unrelated substances or signals interfere, whether samples remain stable and whether results change across operators, instruments or days. Not every property applies to every method. Choose tests that reflect realistic conditions, including challenging inputs, and explain how uncertainty in reference materials affects the conclusions.

#Keep conclusions within the evidence

Report the design, number and distribution of measurements, calculation methods and uncertainty around performance estimates. Investigate failures and invalid outputs rather than silently removing them. Changes to collection, processing or software may require further evaluation, depending on their possible effect on results and the intended application.

Analytical validation does not by itself show that a measurement identifies a clinical condition or improves health outcomes. Those require different evidence. A sound conclusion states what was measured, how well it performed and under which conditions, while identifying untested conditions and important limitations rather than making a broad claim of validity.

#Common misunderstandings

Analytical validation is not the same as showing that a test improves health outcomes. It addresses measurement performance, not whether using the result leads to better care. A method can measure its target reliably while the clinical meaning of that target remains uncertain.

“Validated” also does not mean error-free or suitable for every situation. Evidence from one sample type, instrument or testing setting may not apply to another. Performance can vary across the measurement range, so strong results at higher concentrations do not necessarily establish reliability near a decision threshold.

Another misunderstanding is that detecting a substance means measuring its amount accurately. Detection and quantification are different capabilities. Similarly, repeated results that closely agree can still be consistently shifted away from an accepted reference value.

Finally, analytical validation does not replace ongoing quality checks. Changes in materials, equipment or procedures can affect performance after the initial evaluation has been completed.

#Questions worth asking a clinician

• What exactly does this method measure, in what sample type, and under which collection, storage and testing conditions has it been validated?
• How did you assess repeatability and check for systematic error against an appropriate reference?
• Which performance features, such as detection limits, measurement range and interference, were evaluated for the intended use?
• How much do results vary across operators, instruments, reagent lots and testing sites, and is that variation acceptable for the intended use?
• What separate evidence shows that this analytically validated measurement reflects the clinical condition of interest and improves care decisions?