Four Places Bile Acid Scale-Up Changes the Impurity Profile
Share
Scale-up failures are rarely reported as engineering. They arrive as an impurity that appeared, a crystal form that changed, a yield that dropped twenty points, and a chemistry team looking for a reaction that behaves differently.
Most of the time the reaction is fine. What changed is the physical environment it runs in, and in a bile acid process four of those environments matter more than the others.
One: mixing, and the local concentration problem
In a small flask, adding a reagent over two minutes produces a fairly uniform mixture because the flask is small and the stirrer is proportionally enormous. In a reactor of a thousand litres, the same addition produces a plume of high concentration where the reagent enters, and a region elsewhere that has not met it yet.
Where a reaction is sensitive to stoichiometry, that plume is where side reactions happen. Oxidation steps, selective reductions and anything involving a sterically hindered position are candidates. The impurity that appears after scale-up is often formed in the first thirty seconds of an addition, in a volume that was never uniform.
The control is not a slower addition in the abstract. It is an addition rate set against the reactor's actual mixing time, with the impeller and baffle configuration treated as part of the process rather than as equipment detail.
Two: heat transfer, and the rate that disappears
Heat removal scales with surface area; heat generation scales with volume. A reaction that stays at temperature in a flask because heat escapes through the glass wall will, at scale, sit above its set point for as long as the jacket can only just keep up.
For bile acid chemistry this matters in two places that are easy to overlook.
Exothermic steps. A neutralisation, a hydrolysis or an acidification can release heat faster than a jacket removes it. The result is a temperature excursion during addition, not during the reaction, which is why it is missed in a lab where the excursion is ten seconds long.
Cooling-dependent selectivity. The reduction of a 7-keto intermediate to the desired alcohol is a selectivity question, and selectivity in that step depends on temperature and on the reducing system. A warmer reactor does not fail the reaction. It changes the ratio, and the ratio is the product. The route context is set out in how UDCA is made.
Three: filtration and washing, where the loss is invisible
Filtration in a lab is a Buchner funnel and a visible cake. Filtration at scale is a centrifuge or a filter dryer with a cake whose thickness you cannot see, and a wash that distributes unevenly through it.
Two failures follow. The wash removes less impurity than the lab wash did, because the liquid finds channels through the cake rather than passing uniformly. And the cake retains more mother liquor than expected, which carries dissolved impurities into the next step.
Both show up as a related substance slightly above the historical level, and both are frequently misdiagnosed as a raw material problem. The diagnostic question is simple: did the washing ratio change, and was the cake depth the same as in the lab.
Four: drying, and the gradient that sets the form
Drying at scale is not uniform. The tray or filter dryer has a surface that dries first and a centre that dries last, and the material in between experiences a range of temperatures and solvent activities rather than a single condition.
For a substance with more than one crystal form, that range is a form-selection mechanism. The surface layers may end up in one form and the interior in another, producing material that passes identity and fails a dissolution or a bulk density expectation downstream. The analytical consequences are covered in crystal form and crystallinity.
| Failure | Why the lab did not see it | How it appears at scale | Control |
|---|---|---|---|
| Local concentration during addition | Small volume mixes almost instantly | New impurity, or a higher level of a known one | Addition rate tied to mixing time |
| Heat accumulation | Glass wall dissipates heat | Temperature excursion during addition | Jacket capacity and addition rate co-set |
| Uneven washing | Cake is thin and visible | Related substance drifts upward | Wash volume per cake depth, verified |
| Drying gradient | Thin layer dries evenly | Mixed crystal form, variable bulk density | Temperature ramp and endpoint by content, not time |
What this means for a buyer
Scale-up is where a supplier's process knowledge becomes visible in documents. Four questions expose it.
- Were the commercial batches made on the same equipment train as the validation batches?
- Has the addition rate for the critical step been fixed, or does it vary by campaign?
- Is drying ended on a measured content, or on a time that has worked before?
- When a related substance drifted, what was identified as the cause?
The fourth question is the one that separates a plant that understands its process from a plant that reproduces a recipe. A drift explained by a wash ratio is a controlled process with a known sensitivity. A drift explained as "within specification" is an unexamined one.
Where a process change is being made to fix a scale-up problem, the regulatory route matters as much as the technical fix, and it is covered in rework versus reprocessing.
Primary sources: ICH Q8(R2) for development and design space, the EMA Q11 guideline for process and starting material expectations, and EudraLex Volume 4, Annex 15 for how scale changes are qualified.
Frequently asked questions
Does scale-up always change the impurity profile?
Not always, but often enough that it should be assumed until data shows otherwise. The parameters most likely to change it are mixing, addition rate, temperature control and washing efficiency.
Can a scale-up problem be fixed by changing the solvent?
Sometimes, and it is a significant change rather than a tweak. It alters the impurity purge, the specification basis and the regulatory file, which is why it is usually a last resort rather than a first response.
Why does crystal form change during drying?
Because drying at scale produces a temperature and solvent-activity gradient rather than a single condition. Material at the surface and material at the centre can end in different forms.
Is a lower yield after scale-up a warning sign?
It can be, and it is worth explaining. Yield loss often has a location: mother liquor, wash, or hold-up. An unexplained twenty-point loss usually means material is going somewhere nobody has accounted for.
How many commercial batches are needed to confirm scale-up?
Enough to show the process behaves consistently at the new scale, which is what the qualification exercise is for. Three consecutive batches is the common convention, and the reasoning is set out in our note on process validation.
Does a scale change require regulatory notification?
It can, depending on the market and on whether the change affects the registered process, the specification or the site. The safe assumption is that a scale change is a reportable event until an assessment says otherwise.