Relative Potency (Bioassay)

A small-molecule API's strength can be established by chemical assay — HPLC quantifies exactly how many milligrams of a defined molecule are present, and that number means the same thing in every lab that runs a properly calibrated method. A biologic's potency cannot be established this way, because its activity depends on a specific three-dimensional conformation, post-translational modifications, and functional interactions (receptor binding, enzymatic turnover, cell proliferation or neutralization) that a mass-based assay cannot see — two lots with identical protein concentration by mass can have meaningfully different biological activity if one is partially misfolded, aggregated, or degraded.

The solution the field settled on is to measure potency functionally, in a bioassay (or, increasingly, an orthogonal biochemical surrogate assay), and to express the result not as an absolute unit but as a ratio to a qualified reference standard tested in the same assay run: relative potency = (dose of reference required for a defined response) / (dose of test sample required for the same response), scaled and reported typically as a percentage. This design cancels out a large share of assay-to-assay and day-to-day variability, because the reference and the test sample experience the identical assay conditions, cells, reagents, and operator in the same run.

The three chapters are meant to be read together across an assay's lifecycle: <1032> guides how you design the assay before you have data, <1033> tells you how to prove the assay works (accuracy, precision, range) once you have a validation data set, and <1034> tells you how to analyze every routine run once the assay is in use. A lab that treats these as interchangeable or skips <1034>'s parallelism requirement in favor of simply reading a potency number off a fitted curve is not doing a USP-compliant relative potency determination, even if the arithmetic looks superficially similar.

Two model families dominate relative potency analysis, chosen based on the shape of the underlying dose-response relationship.

1. Parallel-line assays — used when the response is linear against the log-dose over the assay's working range (common for many microbiological and older-style in vivo potency assays). The reference and test sample each produce a straight line when response is plotted against log-dose; if the two lines are statistically parallel (equal slopes), the horizontal distance between them on the log-dose axis is the log of the relative potency.
2. Four-parameter logistic (4PL) models — used for sigmoidal dose-response curves typical of cell-based and ligand-binding bioassays, characterized by four parameters: the lower asymptote, upper asymptote, inflection point (EC50), and slope factor (Hill slope). Relative potency is derived from the horizontal shift between the test sample's fitted curve and the reference standard's fitted curve, again only meaningful if the curves are shown to be parallel (in 4PL terms: equal upper/lower asymptotes and equal slope, differing only in EC50/horizontal position).

A five-parameter logistic (5PL) model, which adds an asymmetry parameter, is sometimes used when the dose-response curve is not symmetric around its inflection point, but 5PL potency calculations are more complex to defend statistically and are used more sparingly, typically only when a 4PL model demonstrably fits the data poorly.

The entire logic of relative potency rests on an assumption: that the test sample and the reference standard contain the same active entity behaving through the same mechanism, differing only in concentration — not in kind. If that assumption holds, diluting the test sample should reproduce the reference's dose-response curve exactly, just shifted along the dose axis. Parallelism testing is the statistical check that this assumption actually holds for the specific run in front of you, and USP <1034> treats it as a mandatory prerequisite, not an optional diagnostic.

• For parallel-line models: a statistical test (commonly an F-test comparing a model that forces equal slopes against one that allows different slopes) checks whether the test and reference regression lines have statistically indistinguishable slopes across the linear portion of the assay.
• For 4PL models: an equivalence-based approach is increasingly favored over a simple significance test — rather than testing whether slopes/asymptotes are 'not significantly different' (which rewards noisy, imprecise assays with easy passes), an equivalence test requires the difference between curve parameters to fall within a pre-specified, scientifically justified equivalence margin.
• If parallelism fails, the assay run is invalid for potency reporting — the two curves are not comparable, and any 'relative potency' number computed from them is not measuring what the test claims to measure. The correct response is to investigate (assay execution error, non-parallel biology such as sample degradation or aggregation) and repeat, not to report the ratio anyway with a caveat.

Because every relative potency result is only meaningful relative to the reference standard used in that run, the reference standard itself is the single most consequential material in the entire potency-testing system, and its qualification and lifecycle management get correspondingly rigorous controls.

1. Primary reference standards are typically established against an international or compendial standard (WHO International Standard, USP Reference Standard, or a company's own in-house primary standard characterized against one of these) with an assigned potency value carrying its own uncertainty.
2. Working reference standards, used for day-to-day routine testing, are qualified against the primary standard through a bridging study — multiple independent assay runs comparing the candidate working standard to the primary, establishing the working standard's assigned potency (often anchored at exactly 100% by definition, with the bridging study transferring the primary's characterized activity onto it) and its acceptable working-life stability profile.
3. When a working reference standard is exhausted and must be replaced, a new bridging study compares the new lot against the current (outgoing) working standard — and ideally is also periodically re-anchored back to the primary or an international standard — so that potency values remain traceable across reference-standard generations rather than drifting cumulatively each time a new lot is bridged only against its immediate predecessor.
4. WHO's guidance on establishing international biological reference standards, and the analogous USP/EP frameworks for compendial reference standards, describe the collaborative study design (multiple independent labs, defined statistical combination of results) used to assign the primary standard's potency in the first place.

Bioassays — whether in vivo animal models, cell-based functional assays, or ligand-binding assays — carry substantially higher inherent variability than chemical assays like HPLC, because they depend on biological systems (cell health and passage number, animal-to-animal variability, receptor expression levels) that are harder to fully standardize than a chromatographic column and detector. It is common and expected for a validated cell-based potency assay's intermediate precision (%CV or geometric CV of the relative potency result across independent runs) to run in the range of 20–30%, sometimes higher, compared to single-digit %CV for a well-run chemical assay.

USP <1033> requires this precision to be characterized explicitly during validation and used to set a scientifically justified specification range — setting a biologic's potency specification with a chemical-assay-style tight tolerance around 100% is a common and serious design error, because it produces a specification the validated assay's own precision cannot reliably meet, generating chronic OOS investigations that are statistical noise rather than genuine product failures.

ICH Q6B, Specifications: Test Procedures and Acceptance Criteria for Biotechnological/Biological Products, is the guideline governing how potency (alongside identity, purity, and other quality attributes) specifications are established and justified for licensed biotech products. Q6B explicitly recognizes that potency for biologics is often assessed through a bioassay reflecting the intended biological activity, and it directs sponsors to justify acceptance criteria using manufacturing experience, stability data, and the assay's own characterized precision — not a default numeric convention imported from small-molecule practice.

In practice this means a biologic's release specification for relative potency (commonly expressed as a percentage range around 100% of label claim, e.g., 70–130% or 80–125% depending on the product and assay) is derived from the same data used to validate the assay under USP <1033> — the specification cannot be tighter than the assay can reliably discriminate, and regulators will ask for that linkage explicitly in a licensing dossier or during inspection.

Because a single bioassay run carries meaningful uncertainty, most potency release testing protocols require combining results from multiple independent assay runs (different days, different analysts, sometimes different plates or animal cohorts) into a single reportable potency value, following the statistical combination rules USP <1034> describes.

• Independent runs are typically combined using a weighted or unweighted mean of the log-potency values (since relative potency is naturally analyzed on a log scale, given its multiplicative/ratio nature), with the combined confidence interval calculated from the between-run and within-run variance components.
• USP <1034> and Ph. Eur. 5.3 provide guidance on testing whether individual assay results are statistically consistent with each other (a form of outlier/heterogeneity check) before pooling them — a run that is a clear statistical outlier relative to the others should be investigated rather than silently averaged in.
• The number of independent runs required to release a lot (commonly two or three) is set during method validation and product specification development, balancing statistical confidence against the practical throughput cost of running a low-throughput, multi-day bioassay repeatedly for every batch.

Because relative potency depends entirely on the reference standard remaining stable and the assay remaining in a state of control, ongoing trend monitoring is a required control, not an optional quality-improvement nicety.

1. System suitability / assay-acceptance criteria are checked on every run (e.g., reference standard curve parameters within historical control limits, positive and negative controls behaving as expected) before any test-sample result from that run is used at all.
2. Levey-Jennings or equivalent control charts track the reference standard's own assigned potency (nominally 100%) run-over-run, catching gradual reference-standard degradation or subtle assay drift long before it would cause an overt specification failure.
3. Statistical process control on product lot potency results over time (see SPC) distinguishes normal assay noise from a genuine shift in manufacturing process potency — a real signal that should trigger investigation rather than being absorbed into 'the assay is just noisy.'
4. Reference standard replacement is scheduled proactively based on stability data and control-chart trend, rather than reactively after a bridging study reveals the outgoing standard has already drifted meaningfully from its assigned value.