Official Source Ledger

Review date: 2026-09-30. This is a source baseline, not a regional applicability decision. The offline helpers make no service/API calls, require no credentials, and have no endpoint, authentication, pagination or response contract to refresh. The collection-citation lookup is separate from the analytical workflow. Retain controlled editions, raw data and calculation outputs.

ICH primary sources

The full public PDFs were downloaded and inspected during this review. Copyright belongs to ICH; these summaries and script tables are adaptations, not ICH-endorsed implementations. ICH permits reuse with acknowledgement under the legal notice in each guideline.

Material corrections from this review:

• M10 LBA selectivity uses ten individual matrix sources, not the chromatographic six (4.2.2). LBA medium QC is near the geometric mean of the calibration range (4.2.4.1).
• Calibration acceptance requires six passing levels, not six attempted levels. Exclusion requires curve refitting/re-evaluation; routine sample/QC bracketing and plate/batch checks remain outside this checker. Anchor exclusion is LBA-specific. The percent pass rules do not themselves establish run acceptability.
• Q2 3.2.3.3 names the SD of intercepts of regression lines. The SE of one intercept is a different statistic and is no longer substituted. Q2 does not provide generic 20% QL confirmation limits.
• Table 2's shared dissolution upper-range cell is 130% of the highest strength's declared content.
• M10 ISR has a study-size minimum; the availability qualification in cross validation is not an ISR exemption. Section 6.2 names bias/agreement methods, not a mandatory TOST equivalence test.

Publisher metadata and public previews only

No subscription standard was accessed. Do not reconstruct the full standard from this skill. Public catalogue scopes/edition metadata are cited below; operational requirements come from the laboratory's authorised controlled text.

USP

• 1220 preview identifies the lifecycle chapter. USP Council of Experts report confirms it became official 1 May 2022. Current revision status still requires USP–NF access.
• 1225 preview explicitly describes a 2025 proposal based on the chapter official from 1 Aug 2017. It proposes Q2(R2) alignment and discusses a proposed 1221 chapter. A DOI page or proposal is not proof of an official effective revision.
• 1226 preview describes selected-characteristic verification for first use; it distinguishes verification from repeating full validation.
• 1224, 1010, 621, 711 and 1092 are pointers only; no adjustment allowances, thresholds or current revision dates are supplied. Verify those in USP–NF.

CLSI

Current publisher pages were read for every listed document. Full standards were not read.

The publisher's May 2026 update discusses modular guides and EP09 revision work. These are not evidence of a new technical edition.

ISO and clinical regulatory context

• ISO/IEC 17025:2017 catalogue: edition 3, published November 2017, current catalogue stage confirmed. Clause labels identify the subject only; full requirements were not read. ISO 15189, ISO 21748 and ISO 5725 remain scope pointers, without asserted editions or detailed requirements.
• CMS CLIA regulation entrypoint and CMS performance-specification interpretation support the distinction between establishment and verification for unmodified FDA-cleared or approved tests. Do not infer US CLIA applicability from CE marking.

Statistical implementation and validation boundaries

These are general statistical methods, not framework-prescribed universal decisions. NIST weighted least squares, NIST weighted fitting, and NIST calibration model checking were checked for variance-model and calibration assumptions. The routines implement t, chi-square and F distributions, weighted linear regression with replicate lack-of-fit, approximate residual runs diagnostics, one-way variance components, Deming/jackknife, a bounded Passing–Bablok subset, Bland–Altman and paired TOST. Tests exercise published quantiles and synthetic numerical cases.

• The scripts are not certified statistical software or a full implementation of any CLSI document.
• Small-sample, normality, independence and variance-model assumptions require review. Grouped accuracy intervals use equal-weight independent group means; a single grouping factor cannot separate confounded day, analyst and instrument effects.
• TOST tests mean-difference equivalence. Individual agreement and decision-point bias need additional pre-stated criteria. Passing–Bablok does not remove all statistical assumptions.
• Synthetic fixtures validate computation, not physical selectivity, robustness, stability, assay performance, instrument integration or regulatory acceptability.