In pharma, a measurement counts once it can be defended. An auditor reads your file, and the spectrum comes after.
A blend finished forty minutes ago waits for the laboratory. A dryer runs to a fixed time, on a margin inherited from a forgotten incident. In both cases the data lands after the decision.
In-line measurement answers that, never alone: reference method, validation plan, access management, audit trail. So the first question is not “which technology” but “which attribute, and what decision does it serve”.
A probe already in place, and out of use? An NIR mounted on a dryer or a blender, qualified at the time, then left without an updated model or an owner. Before buying anything new, the diagnosis: often the hardware is sound, and it is everything else that is missing. What to do with an instrument that has fallen out of use.
Where to start, depending on your role
- You run the plant: start from the operation that waits for the laboratory. Scoping.
- You develop the method: what a proof of concept has to settle. Feasibility.
- You own quality: what software declared 21 CFR Part 11 covers, and what stays with you. 21 CFR Part 11.
- You decide the investment: the cost of a control, calculated on your batches. Cost of a control.
- You want examples: eight real situations, presented without names. Experience.
The three questions that come up most often
Is this blend uniform, and since when
The real question is not whether the batch conforms, but at what point it did. Monitoring during blending gives you a trajectory rather than a verdict, and it sometimes reveals segregation that begins after the endpoint. Watch what the sensor actually sees. Uniformity is measured on increments, not on the batch taken as a whole.
Is this drying finished
A frequent request, and often a profitable one. The gain is calculated on the spread of actual cycle times, not on the cost of the analysis. Two routes: follow a water content, with the calibration that implies, or detect a signal that has stopped moving, which costs far less.
Is this really the product named on the drum
Identification on receipt moves a control from the laboratory to the warehouse. It is the simplest project in the sector: the answer is binary, backed by a library, rather than a value to validate. Its limit is clear. Identifying is not assaying, and conformity against other criteria still has to be established separately.
The dossier stays yours: how far we go with your CMC teams
What we carry. The method, its validation and the pieces that support it: protocol, sampling plan, statistical analysis, validation report, and the measurement’s place in your control strategy. Our experts can also carry the regulatory writing, and most often do so in support of your CMC teams, who keep the filing and the relationship with the authority.
Two routes, and the choice belongs to scoping. In solid dosage the test being replaced is usually a compendial one: uniformity of dosage units, dissolution, assay. You can then demonstrate equivalence with the compendial method, or validate the method on its own merit within ICH Q2(R2) and ICH Q14. Two claims, two volumes of trial work, two calendars, and the choice comes before the trials.
The FDA names the cost of the first route itself: without enough process knowledge, the test-to-test comparison with the conventional method can be the only validation option available, and it can prove burdensome enough to discourage the use of a new technology. PAT Guidance, FDA, 2004.
What the authority looks at. A change of release method falls under the most demanding regime, and the dossier sections involved are known in advance: pharmaceutical development, analytical procedures, validation of those procedures, specifications. The quality summary has to say why the method is fit for its intended purpose, and real time release asks in addition that the control strategy explain what guarantees the attribute once the final test is gone. What validating a multivariate method asks for.
A decision taken with open eyes. Once the method is named for release, going back to the compendial test on a batch that surprises calls for a full documented investigation. That is what separates the level where the measurement steers the process from the level where it releases the batch. The best moment to raise it with the authority is while the dossier is still being built.
Drug product and drug substance: two distinct targets
What follows describes drug product operations, oral solid dosage (OSD) forms and liquid forms alike: granulation, compression, coating, drying, dissolution, emulsification, filling. Upstream, the synthesis and crystallisation of the active substance involve different unit operations, a different regulatory frame — ICH Q11 for development, ICH Q7 for manufacture — and often a different site. The attributes there are reaction conversion, polymorphism and residual moisture, not content uniformity. We give them their own page: drug substance.
When the process runs from synthesis through to the finished product, the two read together: quality secured upstream is quality you no longer have to correct downstream.
Where the measurement sits, operation by operation
Oral solid dosage forms (OSD)
| Unit operation | What is measured | What it changes |
|---|---|---|
| Identity check on receipt | Material identity. Sometimes discrimination between suppliers of the same reference | The control moves to the warehouse, with no destructive sampling. |
| Blending | Active content in successive increments, and its spread | The endpoint becomes a measured state rather than a duration. You also see segregation after stopping. |
| Granulation | Wetting, granule growth, water content of the wet mass | An operation run on operator judgement becomes reproducible between shifts. |
| Fluid-bed drying | Water content, or a drying signature and its endpoint | Stop on the state reached rather than on a fixed time. Watch the gradient: the surface dries before the core. |
| Compression | Content uniformity of the dosage units, identity | First building block of an RTRT file. And the point where sampling becomes central. |
| Coating | Coating thickness and its spread between tablets, by optical coherence tomography | The stop criterion becomes the measured convergence of tablet-to-tablet uniformity. |
| Continuous granulation | Content, moisture, tracking of the material along the line | The batch becomes an interval of time. The measurement is used to divert portions. |
Liquid forms, suspensions and emulsions
Liquid forms do not raise the same questions. There is no powder segregation to watch for, but sedimentation, creaming and coalescence that set in after the operation. And the measurement is usually made through a wall, on a circuit you would rather not open.
| Unit operation | What is measured | What it changes |
|---|---|---|
| Solubilisation and dissolution | Concentration of the active and the excipients during dissolution | The end of dissolution becomes a measured state, not a stirring time. |
| Adjustment and final volume | Final content and homogeneity of the volume | Homogeneity is established in the vessel, without sampling from a circuit you want to keep closed. |
| Dispersion and wet milling | Particle size and its distribution during milling | The stop follows the size reached. Sedimentation after stopping shows too. |
| Emulsification and homogenisation | Globule size, drift towards creaming or coalescence — and phase distribution | The number of passes is decided on the measurement rather than on the recipe. |
| Viscous formulation | Viscosity, during the operation | A rheology deviation shows while it can still be corrected. |
| Filling and unit control | Identity through the glass, by NIR and RAMAN | The control applies to the filled unit, without opening or destroying it. |
The finished-form criteria stay in the laboratory: Ph. Eur. 2.9.40 for uniformity of dosage units, 2.9.19 and 2.9.20 for sub-visible and visible particles, USP <729> for globule size in lipid injectable emulsions. The in-line measurement prepares those controls, it does not replace them.
Each of these lines is decided on your process. Each one assumes a mechanical access point, a reference method, and a calibration burden to settle at scoping.
The guidelines are the work plan. ICH Q8 to Q14 set out Quality by Design, risk management, the process lifecycle, continuous manufacturing and analytical procedure development. ICH Q2(R2) frames their validation. EU GMP and Annex 11, 21 CFR Part 11, GAMP 5 and the ALCOA+ principles cover the data chain and data integrity. USP <905> defines uniformity of dosage units.
Real-time release testing (RTRT) is reachable, and it is built rather than granted: a complete method validation, a demonstrated control strategy and a variation to the dossier. Many projects deliberately aim one step below it, because steering the process without claiming release already delivers most of the gain.
The USP chapter that binds, and the one that explains
The US pharmacopoeia sets near infrared on two levels, and the protocol settles the distinction in one line. USP <856> Near-Infrared Spectroscopy, numbered below 1000, carries the requirements: instrument qualification, procedure, validation and verification. USP <1856> Near-Infrared Spectroscopy — Theory and Practice, above 1000, is informational: <856> itself presents it as a helpful but not mandatory resource. Raman follows the same pair, <858> and <1858>; mid-infrared <854> and <1854>. A validation protocol cites <856>; <1856> helps you write it.
The cost to put in the budget at scoping
Fitting a probe to a blender, a dryer or a press means drilling qualified equipment. Drilling brings requalification, so documents, downtime and verification batches. It is done every day, and it is a line to cost at scoping, before comparing quotes. A budget that carries this line from the start holds all the way to installation. The parts of a control, and the parts of an in-line measurement.
The two questions to settle before purchase
What “21 CFR Part 11 ready” covers, and what your site adds to it
Manufacturers’ declarations of conformity say themselves that overall compliance depends on the user’s procedural controls: training, procedures, account administration. Compliant software is one half. The other half, user access rights, audit trail and model version control, is built at your site, and it is what an auditor comes to see. What a declaration of conformity covers.
Who owns the model once the project is over?
What keeps an installation alive over ten years is the person who reads its residuals and recalibrates it when an excipient supplier changes. Installations that name that person at purchase are the ones still measuring three moves later. So the point to settle before purchase is an organisational one: who owns the model, and with what maintenance budget. What deployment really asks for.
The particular case of CDMOs and CMOs
At a CDMO as at a CMO, the demand changes in nature. Clients increasingly require a Quality by Design approach and in-process measurements, and that requirement comes down through the contract. It is a structural trend in the sector.
The two do not have the same need. A CDMO carries development as well as manufacturing: formulation, stability studies, dosage form development, process validation. In-line measurement serves first to understand the process while it is being built, and it feeds the file the client will submit. A CMO manufactures to a specification already written: measurement serves process control and release, on capabilities it brings itself.
The dividing line is the same in both cases. Instrumenting a capability on a piece of equipment, this dryer can follow a water content, is worth more than a model tied to one client’s product. The first approach is reusable and becomes a commercial argument. The second disappears with the contract. The gain is not the cost of analysis per batch, it is technical transfer time.
Tell us which decision you would rather take sooner. We will tell you whether a measurement brings it back on time.
Forty-five minutes is enough to place the need in your control strategy, to identify the operation where a measurement would pay, and to name the three points that decide it: the reference, the mechanical access, the requalification.