Internal validation and independent evaluation

ICH Q2(R2) distinguishes repeatability, intermediate precision and reproducibility. Repeatability keeps conditions closely similar; intermediate precision considers within-laboratory variation, such as days or analysts; reproducibility examines differences between laboratories.

A favorable result at one level does not automatically establish the others. This classification identifies which variability was examined without dismissing internal work or attributing untested generality to it. Official guideline.

An international study of capillary electrophoresis coupled to mass spectrometry distributed the same batches of samples, reagents and capillaries while participating sites used their own instruments. The design examined transfer across instruments and laboratories without simultaneously changing every material.

Its results distinguished migration times from signal areas, which showed different between-site variability. Merely saying the method was reproducible would conceal that difference between attributes. Primary study.

A multicenter SWATH-MS evaluation combined HEK293 cell material with synthetic peptide series. It examined broad proteomic quantification and characteristics including sensitivity, range and reproducibility.

The materials answered complementary questions: cellular material supplied complexity, while synthetic series enabled controlled comparisons. Observed performance does not certify every matrix or future SWATH-MS implementation. The demonstrated transfer remains tied to the study conditions. Primary study.

An evaluation can include different laboratories while sharing samples, analyses or decisions with the developers. Such collaboration can be useful for testing portability. It should not, however, be described as completely independent of every original component.

To understand independence, identify who performed measurements, set the criteria and processed the data. A different institution adds information, but does not automatically eliminate shared dependencies or guarantee the absence of bias.

If relative times agree while areas differ more, the result points toward a specific quantitative limitation. It does not require declaring the entire platform useless. Similarly, average agreement may conceal components that perform less well.

A transferable report presents criteria, results by attribute, exclusions and shared conditions. It also separates demonstrating precision from demonstrating accuracy: several laboratories can agree without establishing closeness to a reference value. External evaluation is more useful when it explains these differences than when it merely adds a validation label.