This is a ready-to-use SOP for the run-level decision on a relative-potency bioassay or ELISA: is this run valid, is the sample parallel to the reference, and what potency (if any) do you report. It sits downstream of the validated method and upstream of release. Replace every <<FILL: ...>> placeholder and route it through your normal control. A filled specimen follows. This is general guidance to adapt and verify, not legal or regulatory advice.
Document control header
| Field | Entry |
|---|---|
| Document title | Bioassay System Suitability and Parallelism Evaluation |
| Document number | <<FILL: SOP-ID, e.g. SOP-QC-118>> |
| Version | <<FILL>> |
| Effective date | <<FILL>> |
| Supersedes | <<FILL: prior version or "New">> |
| Document owner | <<FILL: role, e.g. QC Bioassay Lead>> |
| Applies to | <<FILL: assay(s) / product(s) in scope>> |
1. Purpose
This procedure defines how <<FILL: COMPANY NAME>> evaluates each run of a validated relative-potency method: acceptance of the curve fits, system suitability, the parallelism equivalence test, calculation of the reportable value, and disposition of the run. The objective is a defensible potency result and a documented, complete record of every run whether it passes or fails.
2. Scope
Applies to relative-potency cell-based bioassays and relative-potency ELISAs governed by validated methods at the sites listed. It does not cover method validation (<<FILL: protocol/SOP ID>>) or PK/immunogenicity ligand-binding assays, which use a different acceptance framework.
3. Responsibilities
| Role | Responsibility |
|---|---|
| Bioassay analyst | Executes the run per the validated method, records raw data, applies the locked analysis, flags suitability or parallelism failures. |
| QC reviewer | Verifies the fit, suitability, parallelism decision, and reportable-value calculation against the record. |
| Biostatistician / SME | Owns the equivalence bounds, variance model, and any statistical questions on a run. |
| Quality Assurance | Confirms no undocumented data exclusion, approves investigations on invalid runs. |
4. Definitions
- System suitability: pre-defined per-run checks (reference EC50 window, signal window, control limits, reference-vs-reference parallelism) that confirm the assay performed as validated before any result is read.
- Parallelism: the demonstration, by equivalence test, that sample and reference dose-response curves have the same shape and differ only by a horizontal shift, which is what makes the potency ratio valid.
- Reportable value: the result format registered in the validated method, usually the geometric mean of N independent assays.
- Equivalence bounds: the data-derived limits, set during validation, that the parallelism confidence interval must fall inside.
5. Procedure
5.1 Fit the curves with the locked settings
- Apply the validated model (4PL or 5PL) and weighting (for example 1/Y^2) exactly as registered. Do not re-select the model or weighting per run.
- Confirm each fit is acceptable: residuals are randomly scattered (no structured pattern near an asymptote), both asymptotes are defined, and the fit diagnostics meet the method’s limits.
- Apply the pre-specified, statistically valid outlier rule only. Do not drop wells or replicates after seeing they hurt the result.
5.2 Evaluate system suitability
Confirm all validated suitability criteria for the run:
| Suitability parameter | Acceptance |
|---|---|
| Reference EC50 within validated window | <<FILL: window>> |
| Signal window (D minus A) | <<FILL: minimum>> |
| Reference-vs-reference parallelism | Passes equivalence bounds |
| Edge / positional controls | <<FILL: within limits>> |
| Control-sample potency (if used) | <<FILL: range>> |
If any suitability criterion fails, the run is invalid. Record it, report no potency, and route to investigation per section 5.5.
5.3 Evaluate parallelism by equivalence test
- Compute the parallelism metric (for example the slope ratio, or the registered combined similarity metric).
- Build the confidence interval (typically 90%) around the sample-to-reference metric.
- Compare the entire interval to the registered equivalence bounds.
- Interval fully inside the bounds: parallel, proceed to the reportable value. Interval breaches either bound: non-parallel, report no potency, route to investigation.
5.4 Calculate the reportable value
- For a parallel run, compute relative potency as the EC50 ratio (reference/sample), accounting for the dilution factor, with its confidence interval.
- Combine independent assays into the registered reportable-value format (for example the geometric mean of
<<FILL: N>>assays). Do not report a single-assay result where the method requires N without a change control.
5.5 Disposition and investigate
- Record every run, passing or failing, with its disposition in the record and the trending log.
- For an invalid run (suitability or parallelism failure), open an OOS/OOT investigation per
<<FILL: SOP-ID>>and look for assignable cause (degraded or aggregated sample, matrix interference, reagent or cell issue, sample potency outside the validated range) before any re-test. - Re-testing to obtain a passing result without an investigation is testing into compliance and is prohibited.
6. Acceptance criteria
A run is valid and reportable when all of:
- Curve fits are acceptable with the locked model and weighting, no undocumented exclusions.
- Every system suitability criterion passed.
- The parallelism confidence interval fell inside the registered equivalence bounds.
- The reportable value was calculated in the registered format.
- The run and its disposition are recorded, and any invalid run has an investigation.
7. References
21 CFR 610.10 (potency) and 600.3(s) (definition of potency). USP <1032> (design), <1033> (validation), <1034> (analysis, parallelism by equivalence), and <111> (bioassay design and analysis). ICH Q2(R2) (validation of analytical procedures) and Q14 (analytical procedure development). ICH Q6B (specifications for biotechnological/biological products).
Describe these compendial chapters in your own words; do not paste their text. Confirm the current version before issue.
8. Record generated: run evaluation record
| Field | Entry |
|---|---|
| Assay / product | <<FILL>> |
| Run ID / date | <<FILL>> |
| Model / weighting (locked) | <<FILL>> |
| Reference EC50 (window) | <<FILL>> |
| Signal window (D-A) | <<FILL>> |
| Suitability | Pass / Fail |
| Parallelism metric and CI | <<FILL>> |
| Equivalence bounds | <<FILL>> |
| Parallelism | Pass / Fail |
| Relative potency and CI (if parallel) | <<FILL>> |
| Reportable value (format) | <<FILL>> |
| Disposition | Valid-reported / Invalid-investigated |
| Investigation reference (if any) | <<FILL>> |
| Analyst / reviewer (name, date) | <<FILL>> |
9. Revision history
| Version | Date | Author | Summary of change |
|---|---|---|---|
<<FILL: 1.0>> | <<FILL>> | <<FILL>> | Initial issue. |
10. Approvals
| Role | Name | Signature | Date |
|---|---|---|---|
| Author | <<FILL>> | ||
| Biostatistics | <<FILL>> | ||
| Reviewer (QA) | <<FILL>> |
Filled specimen
The following shows a completed run evaluation for an example reporter-gene potency assay. Illustrative only.
| Field | Entry |
|---|---|
| Assay / product | Reporter-gene potency assay, Product A |
| Run ID / date | RGA-2606-014, 14 June 2026 |
| Model / weighting (locked) | 4PL, 1/Y^2 |
| Reference EC50 (window) | 4.10 ng/mL (window 3.0-5.5) |
| Signal window (D-A) | 47,150 RLU (min 20,000) |
| Suitability | Pass |
| Parallelism metric and CI | Slope ratio 1.025; 90% CI 0.93-1.13 |
| Equivalence bounds | 0.80-1.25 |
| Parallelism | Pass |
| Relative potency and CI | 94.9% (90% CI 89-101%) |
| Reportable value (format) | Geometric mean of 3 independent assays (this is assay 1 of 3) |
| Disposition | Valid-reported |
| Investigation reference | N/A |
| Analyst / reviewer | L. Chen / P. Adeyemi, 14 June 2026 |
Contrast: on a later run the sample slope came in at 0.78 against the reference’s 1.18, the slope-ratio CI fell to 0.61-0.79 and breached the lower bound of 0.80. That run was dispositioned invalid for potency, recorded in full, and routed to investigation, which found sample aggregation on hold. No potency number was reported and the run was not silently re-tested.
Common inspection findings this SOP prevents
- Testing into compliance: re-running a non-parallel or out-of-spec sample until it passes, reporting only the passing run.
- Post-hoc well or replicate exclusion with no pre-defined outlier rule.
- Using a difference test for parallelism instead of the equivalence test.
- Re-choosing the curve model or weighting per run, or mismatching model between sample and reference.
- Reporting a single-assay result when the validated reportable value is a geometric mean of N.
How to adapt this SOP
- Insert your assay’s validated model, weighting, suitability windows, equivalence bounds, and reportable-value format.
- Point the investigation cross-reference to your real OOS/OOT procedure.
- Attach the pre-specified outlier rule from the validated method.
- Feed the run record into Log: Bioassay/ELISA System Suitability Trending Log.
- Confirm the current version of each compendial chapter and ICH guideline before issue.