Most parenteral manufacturers measure subvisible particles because USP <788> requires it. The data goes into the batch record, the result either passes or fails, and the program moves on. That is a missed opportunity—and, in our experience working with sterile manufacturers across small-molecule, biologic, and combination-product platforms, it is one of the most common gaps in an otherwise mature quality system.
Subvisible particle (SVP) data is not just a release test. Used correctly, it is the earliest, most sensitive indicator you have of mechanical and procedural stress in your process. By the time a visible particle appears in a vial, the upstream event that created it happened days or weeks earlier. SVP trends, monitored deliberately, move long before that visible defect shows up. The question is whether your program is set up to listen to these early warnings.
Why Visible Particle Investigations Take Too Long
When a visible particle is found at release, the typical investigation works backwards through every plausible source: container components, filling line materials, environmental monitoring data, operator handling, hold times. Each thread has to be pursued, ruled in, or ruled out. Meanwhile the batch sits, the supply forecast slips, and the CAPA timeline tightens.
The reason these investigations take weeks is rarely a lack of analytical capability. It is a lack of historical context. If no one was tracking the subvisible signal across the relevant process steps, there is no trend data to point investigators toward the actual root cause. Every investigation starts from zero.
What Gillson Recommends: Treat SVP Data as a Process Sensor
The shift we work with clients to make is conceptual before it is technical. SVP data should be treated like any other in-process sensor—temperature, pressure, conductivity—not like a release attribute. That means:
- Measuring SVP at multiple points in the process, not just at release. Post-formulation, post-fill, post-hold, post-reconstitution—each transition is a candidate measurement point.
- Establishing a process baseline for the SVP distribution your stable process actually produces, including the natural variability.
- Defining action levels on subvisible trends well below the visible particle threshold, so the program can intervene before a visible defect forms.
- Linking SVP changes to specific procedural variables through structured process characterization, so when a trend moves, the likely cause is already mapped.
Where DoE Earns Its Keep
Design of Experiments is the most efficient way to build that map. Manufacturers often resist DoE for SVP work because it feels like a research activity rather than a manufacturing one. We push back on that. A focused DoE varying a handful of procedural variables—hold time, aspiration speed, residual air, agitation, temperature excursion—produces a ranked, statistically defensible understanding of what actually drives SVP formation in your specific product.
The output is not academic. It is a list of process levers prioritized by impact, with the statistical evidence to justify which ones to control tightly and which ones do not warrant the same scrutiny. That ranking becomes a permanent asset for the program—useful for change control, for technology transfer, and for regulatory filings.
The Combination Product Blind Spot
Drug-device combination products are where the lack of SVP-based process intelligence shows up most painfully. Reconstitution introduces hold time. Aspiration introduces shear. Air in the syringe introduces an interface that the protein has to cross. Each of these is a procedural variable the end user controls, not the manufacturer—and each can produce particle formation that the manufacturer is held accountable for.
Manufacturers of combination products who characterize SVP behavior across realistic ranges of these user-controlled variables are far better positioned to defend their inspection results, design their device for low-stress use, and respond to field complaints. Manufacturers who do not characterize this behavior tend to learn about it through post-market surveillance.
What Good Looks Like
A mature SVP program in our experience has four characteristics:
- SVP measurement points are defined in the process based on stress events, not just at release.
- A historical baseline exists for each measurement point, with documented natural variability.
- Trend rules are defined and acted on—drift triggers investigation before a visible defect forms.
- A characterization study, typically DoE-based, maps procedural variables to SVP response so that investigations have a starting hypothesis.
The Bottom Line
The visible particle you investigate next quarter is being created right now by a process variable your current SVP program is not tracking. Moving SVP data from a release attribute to a process sensor changes the economics of the entire visual inspection program—fewer investigations, faster ones when they happen, and a defensible body of evidence that your contamination control strategy is risk-based rather than reactive.
The instruments to do this exist. The methods are established. What is missing in most programs is the decision to use the data this way. That decision is the difference between catching the problem in the vial and preventing it from getting there.
