ICH Q2Validation of Analytical Procedures

ICH Q2(R2) is the harmonised tripartite guideline on the validation of analytical procedures, adopted at Step 4 of the ICH process in November 2023 and superseding the 1994 Q2(R1) text. It defines the technical characteristics a sponsor must evaluate to demonstrate that an analytical procedure is suitable for its intended purpose, and it harmonises terminology across FDA, EMA, PMDA, Health Canada, MHRA, Swissmedic, ANVISA, MFDS, and every other ICH and ICH-observer authority.

The R2 revision restructured the guideline around the analytical-procedure lifecycle established in ICH Q14, recognised multivariate methods (NIR, Raman, chemometric models) explicitly, clarified the validation-characteristics-by-procedure-type table, and aligned the language with the ICH Q8/Q9/Q10/Q11 quality-by-design family.

Not every characteristic applies to every procedure. Q2(R2) Table 1 maps the required characteristics to procedure type: identification methods need specificity only; impurity quantitative methods need all eight except reproducibility; impurity limit tests need specificity and LOD; assay and content uniformity need all eight.

Q2(R1) treated validation as a one-time event preceding routine use. Q2(R2) explicitly embeds validation in the three-stage analytical-procedure lifecycle defined in ICH Q14: (1) Procedure design and development, (2) Procedure performance qualification — which IS the Q2 validation study, and (3) Ongoing procedure performance verification, where system suitability, control charts and periodic re-evaluation prove the method continues to perform.

The lifecycle view introduces the Analytical Target Profile (ATP) — a prospective statement of the performance requirements (e.g. 'assay must report within ±2 % of true value with 95 % confidence over 70–130 % of nominal') — and ties every validation experiment back to whether the ATP is met. This mirrors the QbD-for-process language of Q8(R2) for product.

Process analytical technology (PAT) brought NIR, Raman, FTIR and other multivariate methods to release decisions. Q2(R1) was silent on chemometric models; Q2(R2) addresses them directly. For a multivariate method the eight characteristics still apply, but they are demonstrated through the calibration model: specificity by selectivity in the spectral region, accuracy/precision by cross-validation against reference HPLC, robustness by deliberate model perturbation, range by the calibration set composition.

Lifecycle management is more aggressive than for univariate methods — a multivariate model can drift as raw-material spectral fingerprints shift, and re-validation triggers (new supplier, scale change, model bias drift) must be defined in advance.

1. Method used to release commercial product was validated for R&D scale only — no bridging study after scale-up.
2. Forced-degradation samples (acid, base, oxidative, photolytic, thermal) not included in specificity demonstration.
3. Robustness DoE missing — single-factor-at-a-time experiments do not satisfy Q2(R2) §3.2.7.
4. LOQ established by S/N alone for impurity methods without confirmation of accuracy/precision AT the LOQ.
5. Intermediate precision evaluated by one analyst on two days — Q2(R2) expects analyst AND day AND instrument variation.
6. Range justified by 'assay only' when method also serves content uniformity (requires range to 70 %).
7. System suitability criteria copied from USP without re-derivation against the lab's actual column lots.
8. Re-validation triggered by a method change but the change-control package lacked Q2 characteristic re-evaluation.

• Every LIMS method registry entry carries the ATP, validation summary, validation study reference and re-validation triggers.
• System-suitability acceptance criteria are enforced at the instrument level — failed SST blocks the sequence from releasing results.
• Method version is bound to every result — a method change creates a new version and prior results retain their original method reference.
• Out-of-specification investigations cross-reference the Q2 validation file to test whether the method was within its validated range when the OOS occurred.
• Multivariate model files are versioned alongside the method; bias-drift trending fires re-validation tickets automatically.
• Annual Product Quality Review pulls method-performance trending straight from the same LIMS instance — no separate spreadsheet.