ICH Q2 has governed analytical procedure validation for pharmaceutical registration since the mid-1990s. The revision, Q2(R2), was adopted in November 2023 and took effect in June 2024, alongside a new companion guideline, Q14, covering analytical procedure development.
Between them they moved validation from a one-off study towards a lifecycle. Q14 defines what the procedure needs to achieve, and Q2 is the demonstration that it does.
Q14 introduces the analytical target profile: the performance the procedure must reliably deliver, stated in advance. The ATP matters for the statistics, because it answers the question that undermines most validation work: where the acceptance criteria came from. Under an ATP the criteria are set before the study, from what the measurement has to be capable of.
Which characteristics apply depends on what the procedure is for: an identification test, a quantitative assay, an impurity limit test. The full set, and the analysis each rests on:
| Characteristic | The study, and the guide |
|---|---|
| Specificity / selectivity | Response attributable to the analyte alone: interference and recovery testing |
| Linearity | Proportionality of result to concentration: fitting a calibration curve and assessing linearity |
| Range | The interval over which linearity, accuracy and precision all hold: the measuring interval |
| Accuracy | Closeness to the accepted value, usually by recovery or against a comparative procedure: trueness, precision and accuracy |
| Precision — repeatability | Variation within a run, under constant conditions: the precision components |
| Precision — intermediate | Variation within the laboratory across days, analysts and equipment: establishing precision |
| Precision — reproducibility | Variation between laboratories, from a collaborative study |
| Detection limit | The lowest concentration detectable, not necessarily quantifiable: LoB, LoD and LoQ |
| Quantitation limit | The lowest concentration measurable with acceptable precision: the precision-profile method |
| Robustness | Tolerance of deliberate small changes in method parameters: robustness testing |
Anyone moving between pharmaceutical and clinical laboratory work meets the same quantities wearing different labels. The mismatch causes more confusion than the underlying ideas do.
| ICH Q2 | Clinical laboratory equivalent |
|---|---|
| Detection limit (DL) | Limit of detection (LoD), sitting above a limit of blank |
| Quantitation limit (QL) | Limit of quantitation (LoQ) |
| Intermediate precision | Within-laboratory precision |
| Reproducibility | Reproducibility, between laboratories (the same thing) |
| Specificity | Analytical specificity, interference and cross-reactivity |
| Range | Analytical measuring interval |
| Accuracy | Trueness, where accuracy means trueness and precision combined |
The last row is the one that causes real trouble. In the metrological vocabulary used across the clinical guidelines, accuracy is the combination of trueness and precision, and trueness alone is closeness to the reference value. Reading an ICH accuracy requirement as though it meant that combination will send you looking for the wrong study.
Intermediate precision is defined as intra-laboratory variation, and the definition names its sources: different days, different analysts, different equipment, different environmental conditions. That structure is a nested design, and the analysis separates the variance contributed by each level rather than pooling them into one figure.
Reporting a single standard deviation for intermediate precision discards the breakdown by source. Knowing that between-day variation dominates and between-analyst variation is negligible tells you where to intervene; a pooled figure does not. See the precision components explained for how the levels separate, and establishing precision for the study design.
Q2(R2) is a full revision rather than a tidy-up. The revision clarified the distinction between specificity and selectivity, extended coverage to analytical procedures for biotechnological products, and brought non-linear response models and multivariate calibration models explicitly into scope. The original was written when a straight line was assumed.
Q2(R2) also gives more attention to procedure transfers and to partial revalidation after a change, both of which had been handled inconsistently. Read alongside Q14, the effect is that validation stops being a single event. The ATP defines the target, development evidence can contribute to the validation package, and changes are managed against the same target rather than triggering a full revalidation each time.
Download the multi-laboratory precision example workbook (.xlsx): six samples across three laboratories with five runs and five replicates, separating repeatability, between-run, within-laboratory, between-laboratory and reproducibility components, with a precision profile across the range. It is framed as a clinical protocol, but the nested structure is the one intermediate precision and reproducibility require.
The example workbook is downloading.
It opens in Excel on its own — the data and the finished results are both in it. Analyse-it is what lets you change the analysis and re-run it, try the same study on your own data, or work through it to see how the software handles it.
Every feature from all five editions for 15 days, with no sign-up and no licence key.
Setting acceptance criteria after the study. The ATP exists to prevent this. A criterion chosen once the data are in is not a demonstration of suitability for purpose.
Reading accuracy as trueness plus precision. The vocabularies differ. Check which definition the requirement is using before choosing the study.
Pooling intermediate precision into one number. The nested structure is where the information is. Report the components.
Judging linearity by the correlation coefficient. It will be high for a curved response over a wide range. Read the residuals.
Testing robustness one factor at a time. It cannot find interactions, which are the failures that matter.
Analyse-it produces the statistics the Q2 characteristics rest on, inside Excel:
Every feature from all five editions for 15 days, with no sign-up and no licence key. Model fitting and multi-factor analysis are in every edition, from US$ 155 a year; the precision, linearity and detection capability analyses are in the Method Validation and Ultimate editions, from US$ 475 a year. Validated against NIST and CLSI reference datasets.