Batch to batch variability

Why variation is inherent

Every step from cell bank to vial introduces variation. Cells behave differently between runs. Medium lots differ. Culture duration varies within a permitted window. Harvest timing shifts. Isolation recovers slightly different fractions. None of these is an error; they are the normal tolerances of a process involving biology.

Established biological medicine handles this with a control strategy: defined critical process parameters, in-process controls, release testing against specifications, and stability programmes. The aim is not to eliminate variation but to bound it, and to have shown that variation within the bound does not change clinical performance.

The problem specific to this sector

Bounding variation requires knowing which attributes matter. That in turn requires knowing what the active principle is. In a market where the active principle is generally unidentified, as set out in what is inside an extracellular vesicle, there is no principled basis for choosing which attributes to control.

So control defaults to what is easy to measure: particle count, protein, appearance, sterility. Those are worth measuring. They are also not linked to biological performance by any published relationship, which means a batch passing them is not thereby shown to perform like the last one.

Batches of a cell-derived preparation are equivalent if they meet the same particle count and protein specifications.

Proposed mechanism

Physical equivalence implies biological equivalence.

What has been shown

Comparability in biological medicines is established through defined quality attributes linked where necessary to biological readouts, not through physical measures alone. We are not aware of published work establishing that particle count and protein content predict biological performance for preparations in this market, and the identity of the active principle is generally unestablished.

Highest level reached

Not shown

Main confounders

Physical measures can be met by preparations differing in cargo composition. Cell source variation between banks. Culture protocol drift over time.

GradeNOT SUPPORTED

What would change thisA validated potency assay with a demonstrated relationship between the physical measures and the biological readout, plus published batch data showing the relationship holds across runs. That is the standard comparability exercise, applied here.

What this means for reading a study

Suppose a clinical study of a named product reports a positive result. What does it license?

It licenses a claim about the material used in that study. Extending it to material sold two years later requires a bridge: either that the process has not changed and variation is bounded within demonstrated limits, or that comparability has been shown across the change. Neither bridge is typically documented in this market.

This is not a fringe concern. It is why biological medicines have such extensive comparability requirements when manufacturers change a process, and why regulators treat process changes as significant events rather than internal matters. The underlying logic is that for a biological product, the process is a substantial part of the definition of the product.

Process drift

Beyond run-to-run variation, processes drift. A cell bank is replaced. A medium supplier changes formulation. A step is optimised for yield. Each is a reasonable operational decision and each can change the product.

In a regulated setting, changes of this kind trigger assessment and, where relevant, comparability work. Outside such a setting there is no requirement to notify anyone, and a purchaser may have no way of knowing the product changed. This is the reason we suggest asking suppliers what changed between the last two batches, in why the cell source matters. The value of the question is not the answer so much as whether an answer exists.

What would fix it

The fix is well understood because other sectors have implemented it. Identify a biological readout that correlates with the intended effect. Validate an assay for it. Set an acceptance range. Test every batch. Publish the specification. Handle changes through comparability.

That programme is expensive, and it is difficult when the intended effect is itself not well defined, which is the deeper problem here. A supplier who has begun it has done something substantial. We would list that fact prominently in our listings, and it is one of the few things in this sector that would genuinely distinguish one product from another.

The reader's takeaway

When you see a study cited in support of a product, ask which batches were used and what guarantees the material you would receive resembles them. If the answer is a particle count, the guarantee is physical rather than biological. That is not nothing, and it is not what the citation implies.

Questions readers ask

Why do biological products vary between batches?

Because their manufacture involves living systems whose behaviour varies between runs, along with variation in medium lots, culture timing and harvest conditions. This is inherent rather than a fault, and established biological manufacture controls it rather than eliminating it.

What is comparability?

A formal demonstration that product made after a change, or in a different run, performs equivalently against defined quality attributes and where necessary biological readouts. It is an exercise with evidence attached, not an assertion that two batches look similar.

Does meeting a particle count mean two batches are the same?

No. Physical measures can be met by preparations differing in composition. Establishing equivalence requires attributes linked to biological performance, which requires a potency assay, and those are rarely reported in this market.

Does a clinical study apply to the product I would receive?

It applies directly to the batches used in the study. Extending it requires either evidence that the process is unchanged and variation is bounded, or a comparability demonstration across any change. Neither is typically documented here.

What would a supplier need to do to solve this?

Identify a biological readout related to the intended effect, validate an assay for it, set an acceptance range, test every batch against it, publish the specification, and handle process changes through comparability. That is the standard approach in regulated biological manufacture.