3.3 Standard Operating Procedures (SOPs), Quality Assurance Unit (QAU) & Audit Inspections

SOPs and training as execution controls

DABT Domain I.A.4 and I.B.1 meet in the animal room: a protocol says what unique study to run; standard operating procedures (SOPs) say how the facility always performs recurring tasks. OpenExamPrep independent material for this section treats SOPs as the difference between reconstructable data and folklore. 21 CFR 58.81 requires written SOPs that management is satisfied are adequate to ensure the quality and integrity of the data. The list in 58.81 is not optional trivia — it includes animal care, test and control article handling, laboratory tests, data handling, and equipment maintenance, among other operations.

SOP control means a current, authorized version is the one at the bench, superseded versions are archived in a historical file, and staff can find the method they actually used. Significant changes to established SOPs are authorized in writing by management. That is distinct from a one-time SOP deviation during a study, which the Study Director authorizes and which is documented in the raw data. If every study “deviates” from the same broken SOP, you do not have flexibility — you have an SOP that management has failed to update.

Training is how SOPs become behavior. 21 CFR 58.29 requires that each individual engaged in the conduct or supervision of a study have education, training, and experience, or a combination, to perform assigned functions, and that there be sufficient personnel for timely, proper conduct. Training files should show that the person who mixed diet or ran the HPLC was trained on the current SOP before they generated raw data, not merely that they attended a new-hire orientation years earlier. Personnel with an illness that may adversely affect study integrity are excluded from contact with test systems and articles until the condition is corrected.

QAU duties under 21 CFR 58.35

The Quality Assurance Unit exists to assure management that the study is being conducted in conformance with Part 58. Recap independence from section 3.1: for the study under inspection, QAU is not in the Study Director’s reporting chain and is not among the people directing or conducting that study.

21 CFR 58.35(b) duties a DABT candidate should be able to list without mixing them up with Study Director duties:

1. Maintain a copy of a master schedule of all nonclinical laboratory studies at the facility, indexed by test article, containing test system, nature of study, date initiated, current status, sponsor identity, and Study Director name.
2. Maintain copies of all protocols for studies the unit is responsible for.
3. Inspect each study at intervals adequate to assure integrity, and keep signed records of those inspections. Problems likely to affect integrity go immediately to the Study Director and management.
4. Periodically submit written status reports on each study to management and the Study Director.
5. Determine that no deviations from approved protocols or SOPs were made without proper authorization and documentation.
6. Review the final study report to assure that it accurately describes the methods and SOPs and that the reported results accurately reflect the raw data.
7. Prepare and sign a statement included with the final report specifying the dates inspections were made and the dates findings were reported to management and to the Study Director.

QAU does not approve protocol amendments, does not sign the final report as the person scientifically responsible for the data, and does not authorize SOP deviations in place of the Study Director. Those remain Study Director (and, for SOP text changes, management) functions.

Written QAU procedures and the method of indexing QAU records are themselves controlled documents. FDA, under 21 CFR 58.35(d), can inspect the fact of inspections (dates, study, phase or segment, inspector name). Treat narrative finding memos as internal compliance tools for management and the Study Director.

Critical-phase inspections versus facility inspections versus protocol/report audits

Name the tool by what it looks at.

A critical-phase inspection that only reads the dosing binder after lunch is a paperwork review, not a critical-phase inspection. A report audit that never opens the raw-data archive cannot confirm that reported means match original observations. OECD quality-assurance guidance uses similar buckets (study-based, process-based, facility-based) for multi-site work; lead QA still inspects the assembled report including test-site contributions.

Method validation: full versus fit-for-purpose

Analytical and biomarker methods need documented performance that matches the decision the method supports.

• Accuracy: closeness of measured values to a known true or accepted value (spike recovery, certified reference).
• Precision: scatter among repeats (repeatability within a run; intermediate precision across days, analysts, or instruments).
• Specificity (selectivity): ability to distinguish the analyte from matrix, vehicle, metabolites, and co-medications.
• Sensitivity: ability to detect small amounts; often discussed with limit of detection (LOD) and limit of quantitation (LOQ) or the bioanalytical lower limit of quantitation.
• LOD is the lowest level distinguishable from blank; LOQ is the lowest level that can be quantified with acceptable accuracy and precision. Reporting a concentration below LOQ as if it were a precise dose is a data-integrity problem.

Full validation (ICH Q2 for many quality-control assays; ICH M10 / FDA bioanalytical guidance for concentration methods supporting regulated pharmacokinetic or toxicokinetic decisions) is expected when the number will be used as a primary regulated result — formulated-diet assay, toxicokinetic plasma concentrations, clinical pathology analyzers used as study endpoints.

Fit-for-purpose qualification is appropriate for many exploratory biomarkers, early method development, or investigative endpoints that will be labeled as supportive. Fit-for-purpose still needs a written rationale, some demonstration that the method measures what you claim, and transparent limitations. It is not a synonym for “unvalidated and unwritten.” Conversely, a fully validated HPLC method used outside its validated matrix (plasma method applied to 24-hour diet extracts without additional work) is no longer validated for that use.

Alternative methods: OECD-adopted tools, limits, and omics

Animal-reduction policy and chemical-registration needs have produced OECD Test Guidelines that DABT candidates should recognize by number and by what decision they can support.

Skin sensitization. The adverse outcome pathway starts with covalent binding to proteins (key event 1), keratinocyte activation (key event 2), and dendritic-cell activation (key event 3). OECD TG 442C covers in chemico peptide-binding methods such as the Direct Peptide Reactivity Assay. TG 442D covers in vitro keratinocyte reporter methods (for example KeratinoSens). TG 442E covers in vitro dendritic-cell activation methods (for example h-CLAT). OECD TG 497 describes defined approaches: specified information sources combined with a fixed data interpretation procedure so that the same inputs yield the same hazard or potency call. Defined approaches exist to overcome limitations of any single assay; they are not a license to ignore applicability domains (poorly soluble chemicals, complex mixtures, pro-haptens that need metabolism the in vitro system lacks).

Eye irritation / serious eye damage. TG 437 (bovine corneal opacity and permeability, BCOP) and TG 438 (isolated chicken eye) can identify many UN GHS Category 1 chemicals and many chemicals that need no classification, but they are not complete stand-alone replacements for every Category 2 call. BCOP has a documented tendency to over-predict (false positives) some alcohols and ketones as serious eye damage and to under-predict (false negatives) some solids. TG 492 reconstructed human cornea-like epithelium methods are widely used to identify chemicals not requiring classification; an inconclusive or irritating result still needs a strategy, not a forced Category 1 label.

Phototoxicity. OECD TG 432 (3T3 Neutral Red Uptake) compares cytotoxicity in BALB/c 3T3 fibroblasts with and without UVA. A positive predicts phototoxic potential (photoirritation). It does not replace an in vivo photocarcinogenicity bioassay, and it does not evaluate photoallergy. Highly colored or poorly soluble chemicals can challenge the assay; some UV-absorbing chemicals produce in vitro positives that do not translate to in vivo photoirritation.

Advantages of these methods: shorter timelines, less animal use, often better intra-laboratory control of exposure, and for TG 497 a transparent combination rule. Limitations: restricted applicability domains, metabolic competence gaps, mixture/formulation problems, and known false-positive or false-negative patterns that force a weight-of-evidence or additional test rather than a single-assay slogan.

Omics (transcriptomics, proteomics, metabolomics) can generate mechanistic hypotheses and support an adverse outcome pathway narrative. Unless the marker has been qualified for a defined context of use (for example a qualified hepatic injury biomarker with known performance against histopathology), treat the heatmap as hypothesis-generating. Hundreds of differentially expressed genes at a high dose that also produced necrosis do not, by themselves, become a GLP-qualified replacement for histopathology, and they do not automatically prove that a histopathology finding was a false positive.

Keep SOPs, QAU inspections, validated methods, and alternative assays in one mental model: each is a control on data quality. None of them, including GLP itself, replaces scientific judgment about whether the test is fit for the regulatory question.