The Scale-Up Failure Hiding in Plain Sight
A commercial liquid capsule line can use the same formula, shell, and nominal machine settings as a successful pilot run yet produce fill-weight drift, leaking seams, or assay variation. The recurring cause is often not a defective pump. It is loss of control over the formulation's physical state between the compounding vessel and the capsule.
A process map for liquid capsule manufacturing may present compounding, dosing, sealing, drying, and inspection as separate operations. At commercial scale, they behave like one coupled system. The temperature and flow history of the fill influences pump pressure; pump pressure influences dose and air entrapment; dose and air influence seam formation and drying. A problem can therefore appear at inspection even though the initiating variable changed much earlier.
The useful scale-up question is not whether the commercial machine can run faster. It is whether the fill reaches every dosing point in the same physical state as the state used to qualify the formulation.
That physical state is best described through viscosity, temperature, density, air content, and, for suspensions, particle distribution. Among those variables, viscosity is the hidden connector. It changes rapidly with temperature and shear, and it translates directly into flow resistance, pump response, cutoff behavior, and seal cleanliness.
A Liquid Fill Has a Process State, Not a Single Specification
Viscosity is often recorded as one batch result, such as a value measured at room temperature after compounding. That is useful for release, but it is not enough for scale-up. A fill can meet that result and still behave differently at the nozzle because dosing occurs at another temperature, after another hold time, through a different amount of tubing and filtration.
For a relatively simple oil solution, viscosity may fall as product temperature rises. For a lipid and surfactant blend, the change may be nonlinear near a phase transition. A suspension can show shear-thinning behavior: it flows easily under agitation but becomes much more resistant after standing. A formulation containing dissolved gas or entrained air may look within viscosity limits while responding unpredictably to a positive-displacement pump.
A specification such as 800 mPa·s at 25°C does not define what enters a 40°C dosing head. It does not describe the pressure required after a filter, the response of a nozzle at production speed, or the way the fill behaves after two hours of recirculation.
For that reason, the right development question is not What is the viscosity of the batch? It is:
- What is the viscosity at each relevant temperature?
- How does it change at the shear rates created by the mixer, pump, filter, and nozzle?
- Does the material change after the maximum expected hold time?
- Does de-aeration alter its apparent flow behavior?
- For a suspension, does the particle concentration remain uniform from the top of the vessel to the dosing head?
Viscosity becomes a critical process variable when the answers affect fill weight, content uniformity, seal integrity, or dissolution. That is the point at which a simple laboratory test needs to become a scale-up control strategy.
Why Pilot Production Hides the Problem
Thermal inertia changes with vessel size
A small compounding vessel has a favorable surface-area-to-volume ratio. The jacket, impeller, and surrounding air can bring most of the batch close to one temperature quickly. A larger vessel has more material in the center and proportionally less heat-transfer area. A single temperature probe can report the setpoint while a temperature gradient remains between the wall, core, outlet, and recirculation loop.
The problem continues after the fill leaves the tank. Filters, valves, transfer hoses, manifolds, and dosing heads all exchange heat with the room. An insulated, short pilot line may deliver nearly the same product temperature from tank to nozzle. A commercial line with multiple heads and longer paths may deliver a different temperature to each head, especially during start-up, low tank level, or a production pause.
For a formulation with a steep viscosity-temperature curve, a few degrees can move the process from a stable dosing range to a high-pressure, slow-flow condition. The tank display can remain green while the product at the point of use is already outside the window established during development.
Residence time changes with batch size
The amount of time a fill spends in the system is a process parameter, not a scheduling detail. A 20-liter batch may be consumed in a short pilot run. A 500-liter batch may remain under agitation, recirculation, heating, and exposure to air for hours.
The relevant calculation is simple:
Turnover time = working volume ÷ transfer flow rate.
If the commercial process does not increase transfer capacity in proportion to batch volume, turnover time grows. If transfer capacity is increased aggressively, pressure drop and shear may grow instead. Both changes can alter the fill.
Longer hold time can lead to:
- settling in a suspension
- crystallization or precipitation after local cooling
- solvent or water loss at the vessel surface
- oxidation or foam formation
- changes in surfactant structure
- viscosity drift after repeated recirculation
A fill that is uniform at the end of compounding may not be uniform at the last dosing head. A formulation that is clear when warm may develop crystals in a cool section of hose. Those are not separate scale-up mysteries; they are consequences of a different material history.
Mechanical energy and air entrainment change
Commercial pumps and transfer loops often move more material through filters, elbows, valves, and narrow manifolds. That adds pressure drop and can introduce more shear than the pilot setup. High shear may be beneficial for dispersion, but it can also heat the product, break suspension structure, or create foam in a surfactant-rich fill.
Air is particularly disruptive because it is compressible. A pump chamber containing a small amount of air does not respond like a chamber filled entirely with liquid. The stroke can be mechanically correct while the delivered dose varies. In a softgel, bubbles can become visible defects or weak regions in the fill. In a liquid-filled hard capsule, air can cause underweight units, inconsistent cutoff, and pressure fluctuations at the nozzle.
Vacuum de-aeration must therefore be verified at the commercial batch size and after the final high-shear step. A beaker that looks bubble-free does not prove that the production vessel, return line, and dosing manifold are equally free of air.
Throughput turns one control loop into several
At low speed, operators have time to adjust temperature, pressure, and pump settings before a small deviation affects many units. At commercial speed, the same deviation can produce thousands of capsules before a manual check identifies the trend.
Multiple dosing heads introduce another layer. The first head may be close to the heated manifold; the last may have a longer path or a different pressure balance. A single average fill weight can hide head-to-head differences. Rotary-die softgel equipment adds another synchronization challenge: the fill pump, injection wedge, ribbon temperature, die speed, and ribbon thickness must remain coordinated as line speed changes.
Scale-up by simply increasing revolutions per minute is therefore unreliable. Matching a nominal agitator speed or pump setting does not guarantee matching heat transfer, turnover, shear, or material state.
How the Hidden Variable Becomes a Visible Defect
Fill-weight drift is often a rheology signal
The basic relationship is straightforward:
Fill mass = metered volume × density.
A positive-displacement pump improves repeatability, but it does not make viscosity irrelevant. The ideal stroke volume may be fixed, while actual delivery changes because of suction filling, valve response, internal slip, tubing expansion, nozzle restriction, or cutoff delay. Higher viscosity usually increases pressure demand and can change how completely the pump chamber fills. Lower viscosity can increase stringing, dripping, or backflow after the stroke.
A declining fill-weight trend across a run is often interpreted as pump wear or a calibration problem. Those are possible causes, but a product-temperature trend or rising pump pressure can point to viscosity drift instead. If every dosing head moves in the same direction, the shared fill path, thermal state, or bulk composition deserves priority. If only one head changes, a local restriction, nozzle, or calibration issue is more likely.
The average is not enough. A batch can have an acceptable overall mean while the first, middle, and last portions show a clear slope. Time-stratified data exposes the mechanism that a composite sample hides.
A bubble can look like a dosing problem
Entrained air creates an apparent volume problem before it creates a visible defect. The pump compresses the bubble, the capsule receives less liquid than expected, and the next stroke may deliver a different amount when the bubble moves or collapses. A balance sees the consequence but not the cause.
In softgels, visible bubbles, dimples, or weak seams can be related to air in either the fill or the shell mass. In liquid-filled hard capsules, air can produce variable fill weights and unstable dosing pressure even when the shell bodies are dimensionally consistent.
A useful diagnostic is to compare fill-weight variation with pump pressure and product temperature. If weight variation grows when pressure oscillates, inspect the fill path for air, foam, restriction, or viscosity change before changing the nominal stroke volume.
Leaks are a flow problem before they are a shell problem
A leak is often blamed on defective gelatin, a poor capsule shell, or inadequate drying. Those causes exist, but the fill's flow behavior at the sealing interface is equally important.
In a softgel, the fill and two ribbons meet at the die wedge under a narrow timing and temperature window. If the fill becomes too viscous, injection pressure can rise and force material toward the seam. If it becomes too fluid, it can spread or wick into the seal area before the ribbons close. The exact defect pattern depends on the formulation and die geometry, but both sides of the viscosity window can damage seal integrity.
In a liquid-filled hard capsule, the body and cap may close correctly while residual liquid remains on the joint. That residue can prevent the banding solution from wetting the shell uniformly. A low-viscosity fill may migrate through the joint before sealing; a high-viscosity fill may string from the nozzle or leave a heavy deposit on the rim. The capsule can have the correct fill weight and still fail the leak test.
The banding solution introduces a second rheology problem. Its solids content, temperature, application volume, and drying behavior must be controlled separately from the fill. Replacing a gelatin shell with HPMC also changes the wetting and sealing behavior; the same nominal banding settings cannot be assumed to transfer.
Assay and dissolution failures can begin during the hold
Not every scale-up failure appears as an underweight capsule. A suspension can deliver the correct average weight while the active concentration varies from one portion of the batch to another. A cold section of line can cause a dissolved ingredient to precipitate, creating crystals that dissolve differently from the original formulation. Extended heat or oxygen exposure can change the fill chemistry before encapsulation begins.
Shell-fill interactions can also be affected by thermal and moisture history. The capsule may pass initial visual inspection and fill-weight testing, then show a dissolution shift during stability because the shell has softened, dried, or interacted with the fill differently than it did in the pilot process.
The common thread is not that viscosity causes every failure directly. It is that viscosity is a sensitive indicator of whether the fill has followed the intended process history. A change in viscosity can be the first measurable sign of a broader change in composition, temperature, dispersion, or air content.
Softgel and LFHC Lines Move the Risk, Not Remove It
Softgel and liquid-filled hard capsule production expose the same hidden variable in different places.
Softgel manufacturing combines shell formation, fill injection, cutting, and sealing in one event. The gel mass must form a ribbon with consistent thickness and moisture while the fill arrives at the wedge with the right temperature and flow resistance. The process window is coupled: changing fill temperature can affect injection pressure and seal behavior, while changing ribbon conditions alters how much pressure the seam can tolerate.
Liquid-filled hard capsule production starts with a preformed shell, so it separates shell manufacturing from liquid dosing. That can simplify some formulation changes, but it adds a distinct banding operation. The fill must dose and cut off cleanly, the capsule joint must remain clean, and the banding solution must form a continuous seal and dry at the required rate.
The machinery is different, but the control logic is the same: define and control the material state at the point where the fill meets the shell. A process that controls only the tank temperature or only the final leak result is controlling too late.
The Corrective Strategy: Control the Material State at the Point of Use
1. Replace one viscosity number with a process map
Measure the fill across the temperature range expected during start-up, steady production, pauses, and shutdown. Use the actual instrument, spindle or geometry, shear rate, sample preparation, and sample age that will define the manufacturing method.
For non-Newtonian fills, a single rotational reading can be misleading. Measure at shear conditions that resemble mixing, pumping, and dosing. For suspensions, add yield behavior, particle-size distribution, settling rate, and concentration uniformity after the longest expected hold.
The output should be a performance-linked dosing window, not an arbitrary viscosity range. The window should identify the conditions that produce acceptable:
- fill-weight mean and coefficient of variation
- pump pressure and stroke response
- cutoff or nozzle cleanliness
- seal integrity
- assay and content uniformity
- dissolution behavior
If a formulation performs acceptably only at one temperature in a narrow laboratory test, that limitation needs to be visible before commercial equipment is ordered.
2. Map the thermal path from tank to capsule
Install or verify temperature measurement at the bulk vessel outlet, after critical filters, near the manifold, and at representative dosing heads. For a softgel, include the fill line and the injection wedge area. For an LFHC line, include the fill nozzle and banding reservoir.
Record the temperature during:
- initial warm-up
- the first filled capsules
- steady state
- low tank level
- a planned line stop
- restart after the stop
- the final portion of the batch
The goal is to find the cold and hot spots that a tank probe cannot see. Insulated or heat-traced lines, properly sized jackets, reduced dead legs, and controlled recirculation can narrow the temperature spread. A start-up interlock that prevents dosing until the line reaches equilibrium is often more effective than allowing operators to correct the first several minutes manually.
Temperature sensors also need calibration and sensible placement. A precise sensor in a nonrepresentative location produces precise but irrelevant data.
3. Calibrate the pump with the real product and real path
Pump calibration should use the actual formulation, commercial tubing or piping, final filter, nozzle, and dosing temperature. Calibration with water or a low-viscosity surrogate can confirm that the pump turns, but it cannot predict how the production fill will behave.
A practical study collects a defined number of strokes gravimetrically at low, nominal, and high points of the expected viscosity window. Repeat the check at the beginning, middle, and end of the run, and across all dosing heads where applicable. Record pressure at the same time.
Checkweighing should be treated as a trend-control system rather than a pass-or-fail ritual. An alarm on a growing slope can stop a systematic defect early. A single average taken after the run cannot recover capsules made while the process was drifting.
4. Design mixing, de-aeration, and hold time together
The mixing system must keep a suspension uniform without creating a vortex that entrains air. The transfer system must provide enough movement to prevent settling without exposing the fill to unnecessary shear and heat. Those requirements can conflict, which is why simply increasing agitator speed is rarely a complete solution.
Set a maximum bulk hold time based on data from the actual formulation. Include the time before filling, the time in recirculation, and the time during any pause. Define what happens after a stop: whether agitation continues, whether the line is drained or recirculated, and how the first post-restart capsules are handled.
For a suspension, sample from representative locations or through a validated recirculation loop and compare active concentration, density, and viscosity. For a solution or SMEDDS fill, check clarity, precipitation, phase behavior, and viscosity after the maximum thermal exposure.
Increasing viscosity to slow sedimentation can make dosing and sealing worse. The better solution may be a different particle-size distribution, a controlled suspension structure, a gentler agitation pattern, or a shorter hold.
5. Link fill controls to shell and seal controls
For softgels, monitor gel mass viscosity, ribbon thickness, ribbon moisture, fill temperature, injection pressure, wedge temperature, and die speed as a connected set. A stable fill cannot compensate for a ribbon that is too dry, too thin, or out of thermal balance with the wedge. Samples should represent the die width and the production timeline, not only the center of the run.
For liquid-filled hard capsules, control fill temperature and viscosity alongside nozzle cutoff, capsule-joint cleanliness, banding solution viscosity, application volume, band continuity, and drying conditions. A band solution that is too thin may fail to build a continuous barrier; one that is too thick may produce an uneven band or fail to dry before handling.
The acceptance test should follow the failure mechanism. Visual inspection can find obvious leaks and cosmetic defects, but pressure, vacuum, dye, or other validated leak tests are needed when seal integrity is a critical attribute. Dissolution and stability testing remain necessary because a seal that looks sound on day one can behave differently after storage.
6. Validate transitions, not only steady state
Commercial failures frequently occur at process edges:
- cold start before the line reaches temperature
- the first capsules after a pause
- low tank level and reduced suction head
- the last portion of a batch
- maximum line speed
- minimum line speed
- maximum bulk hold time
- a change from one dosing head to another
- high or low room temperature and humidity
A credible performance qualification challenges these conditions within safe operating limits. Samples should be time-stratified and, when relevant, head-stratified. Evaluate individual weights, weight trends, assay or content uniformity, leak rate, band or seam integrity, moisture, and dissolution.
Three nominal batches that all run under ideal conditions demonstrate repeatability at the center of the window. They do not demonstrate robustness at the edges. Robustness comes from showing that a controlled change in temperature, viscosity, speed, or hold time produces a predictable response and remains within the established acceptance limits.
Reading the Process Data Like a Root-Cause Investigation
The defect pattern usually contains more information than the defect name.
- All heads drift together: investigate bulk temperature, viscosity, density, common pressure, hold time, and recirculation before recalibrating individual nozzles.
- One head is consistently low or high: inspect that head's filter, tubing, valve, nozzle, pressure balance, and calibration.
- Weight shifts after a pause and returns to normal later: look for line thermal equilibration, air movement, or product settling.
- Pump pressure rises while output falls: suspect increasing viscosity, cooling, a restriction, or a suspension deposit. The direction of the weight change depends on pump design.
- Average fill weight passes but leaks increase: examine fill behavior at the seam, shell moisture, ribbon condition, joint cleanliness, banding solution, and drying rather than assuming the dosing volume is correct enough.
- Top, middle, and bottom samples show an assay gradient: investigate settling, inadequate turnover, or a nonrepresentative sampling method.
- Initial release passes but dissolution changes on stability: examine precipitation, shell-fill compatibility, moisture exchange, thermal exposure, and any change in the fill's physical state during storage.
Short, controlled engineering trials can separate causes. Hold pump settings constant while deliberately comparing two qualified product temperatures, then compare pressure, weight, and seal results. Repeat with the dosing head temperature measured independently from the tank. If output follows the product state rather than the nominal pump setting, the scale-up problem is rheological or thermal. If output does not move with product state but changes by head, the investigation can shift toward equipment balance.
What a Transferable Scale-Up Package Contains
A formulation is not ready for commercial transfer merely because it can be encapsulated on a pilot machine. The technical package should contain evidence that the material state survives the full path.
At minimum, it should include:
- A viscosity-temperature profile and, where necessary, a shear-rate profile.
- Hold-time data covering the longest expected compounding, recirculation, and pause periods.
- A thermal map from the vessel to the dosing point, including start-up and restart.
- Gravimetric pump and nozzle performance at actual product conditions.
- A de-aeration study at commercial batch size, with evidence from the dosing point.
- Suspension uniformity data or solution stability data across the batch timeline.
- A linked softgel ribbon and wedge study or an LFHC fill-and-banding study.
- Time-based and head-based in-process sampling with trend limits.
- A response plan for temperature, pressure, weight, and leak excursions.
The package should also identify which changes require re-evaluation. A new filter, longer hose, different shell material, altered surfactant level, revised batch size, or higher line speed can change the physical state even when the ingredient list is unchanged.
The Cost of Treating Scale-Up as a Speed Increase
End-of-line inspection is an expensive substitute for upstream control. In a hypothetical 500,000-capsule campaign, a 2% systematic defect rate creates 10,000 affected units before sorting, investigation, and rework. If the defect is caused by a gradual temperature or viscosity drift and sampling occurs only at the end, a small number of tested units may not reveal how long the process was out of control.
The better economic decision is to detect the drift where it begins: product temperature at the dosing head, pressure across the fill path, weight by time and head, and seal performance during the run. A process that stops after a defined alert can preserve material. A process that waits for final inspection turns a recoverable control issue into a batch disposition problem.
Commercial equipment is not inherently less capable than pilot equipment. It simply exposes variables that pilot production often allows to remain hidden. The durable fix is to transfer a physical state and a validated process window, not just a formula, rpm setting, and tank temperature.
When the fill reaches the nozzle or die wedge at the intended temperature, viscosity, air content, and composition, pump calibration becomes meaningful, sealing becomes predictable, and scale-up becomes an engineering exercise rather than a search for the next defect.