“Does my weight gain tell me what happens tablet by tablet? What is the gap between the best and the least coated? Will my functional coating hold its performance?” Three questions that steering on weight gain leaves open, because a mean does not describe a dispersion.
That gap is what decides. On a cosmetic coating it shows. On a functional coating, enteric, modified release, moisture barrier, it sets the performance of the product.
The microscope cross-section remains the reference: a few tablets taken at the end of the cycle, sectioned, observed, and it is that measurement which gives the value that counts. What so small a number of units cannot give is the dispersion of the batch.
What a coating operation asks for sits between the two: the distribution of thicknesses, while the pan is turning.
In short
Optical coherence tomography (OCT) measures thickness tablet by tablet, without contact and with no model to calibrate: it gives the spread, not just an average. NIR remains relevant when the coating is too pigmented for optics; it then follows coating progress by correlation, with a model to maintain.
What is actually measured
A local thickness, on one tablet, at one instant
OCT produces a cross-sectional image of the coating and measures the distance between the surface of the coating and the coating to core interface. A direct geometric measurement at the micron scale, rather than an estimate of a correlated property.
The dispersion between tablets is the useful information
Measuring thousands of tablets during the cycle gives a distribution that widens, then tightens as the coating evens out. The convergence of that dispersion is an objective endpoint criterion, bearing on uniformity rather than on the quantity sprayed.
Surface roughness counts as much as thickness
The same measurement gives the surface state of the coating, which is a descriptor in its own right. Two batches of the same mean thickness and different roughness behave differently, at dissolution as at printing.
The quantity comes out of the physics of the measurement: there is no chemometric model to build, to validate or to keep alive. The instrument returns an optical path, converted into a physical thickness by the refractive index of the coating. No reference sample campaign, no design of experiments, no revalidation at each change of excipient batch.
There is still an entry price, and it sits elsewhere. In the geometry rather than in the calibration, and that is where the feasibility of a coating project is decided. Why some measurements call for a model and others do not.
Which technology for which case
| Technology | What it follows | Suitable matrices | Calibration burden | Installation | Known condition |
|---|---|---|---|---|---|
| Optical coherence tomography OCT | Thickness tablet by tablet, dispersion, surface roughness | Lightly to moderately pigmented coatings, functional coatings | None: direct geometric measurement | Contact-free, measurement head in the pan wall or in-line | Probe to sample distance held to within a few tens of microns. Heavily pigmented coatings are the demanding case |
| NIR spectroscopy | Progress of the coating, by correlation | Tablets in a pan | Full, with upkeep of the model | Probe in the pan wall | Also responds to coating composition, residual moisture content and compaction |
On coating, optical coherence tomography returns a thickness rather than a signal correlated with it. NIR keeps its place where the coating is too pigmented for coherent optics, or where the stake is the progress rather than the uniformity.
The factor that decides these projects
The distance between the measurement head and the tablet decides ahead of resolution, rate or software.
The output of the measurement depends on that distance, and the literature characterises it explicitly before any installation: it is a parameter that gets quantified and whose admissible window is fixed, product by product. The mechanical eccentricity of a coating pan is counted in hundreds of microns.
The gap is therefore an order of magnitude, and it is handled: head positioning, mount design, distance monitoring and correction. It is handled at the installation study rather than at commissioning, which is why the eccentricity of the pan is measured before the instrument is ordered.
The best question to put to a supplier in a meeting: what working distance tolerance, and how do you hold it on a real pan?
What is worth preparing on your side
- The composition of the coating and its pigment load. The first criterion of applicability, ahead of everything else. A coating heavily loaded with titanium dioxide or iron oxides changes the nature of the problem.
- The geometry of the pan: diameter, speed, measured eccentricity, and the optical access available. An eccentricity measurement taken before the project is worth several meetings.
- Reference tablets whose thickness has been established by an independent method, section and microscopy. They serve to verify rather than to calibrate, and that verification stays essential.
- The history of durations and weight gains, batch by batch. It is what puts a figure on the gain before any investment, since the recoverable margin is the spread of those durations.
- Deliberately perturbed batches, where the material and the regulation allow it. They are what establishes that the measurement sees a real drift rather than a nominal trajectory.
Conditions for success and limits
Heavily pigmented coatings mask the interface
The physical condition of the method. Titanium dioxide and iron oxides scatter the light to the point where the coating to core interface stops being detectable. No pigment threshold is published — what the literature provides is a binary catalogue: of 22 coating formulations put through OCT, six only let the coating to core interface be seen, and none of those containing titanium dioxide. Hence the authors’ standing recommendation: an optical transparency trial on your formulation, before any commitment.
A way round exists: scattering descriptors obtained by unsupervised machine learning on the images, correlated with tablet thickness and so, indirectly, with coating thickness. They were established on one formulation, loaded above 30 per cent titanium dioxide, and the authors themselves present the extension to other scattering coatings as an expectation rather than a result. The measurement there becomes indirect and referenced: on a heavily pigmented coating the method trades away its main advantage, which is worth knowing before a project is built on it.
Catalogue: Lin, Dong, Markl, Zhang, Shen, Zeitler, Journal of Pharmaceutical Sciences 106(10), 2017, doi:10.1016/j.xphs.2017.05.032. Scattering descriptors: Fink, Gartshein, Khinast, Journal of Pharmaceutical Sciences 113(6), 2024, doi:10.1016/j.xphs.2024.01.008.
TERAHERTZ answers a different question here
TERAHERTZ measures multilayer thicknesses very well, tablets included: validation against microscopy of cross-sections dates from 2007, validation against X-ray microtomography from 2012, and an in-line sensor was measuring the coating of individual tablets in real time in a coating pan as early as 2011, with no chemometric model. What is missing is therefore not the evidence but the resolution: terahertz pulsed imaging works from about 40 µm, where OCT reaches down to 10 µm. The two are complementary, and the reference literature says so — OCT from 10 to 60 µm, terahertz from 40 µm and above, indifferent to pigments where OCT loses ground.
Lin, Zhang, Markl, Zeitler, Shen, Applied Sciences 2018, 8(12), 2700, doi:10.3390/app8122700, open access. What we offer is a different use of TERAHERTZ: passing through packaging with no radiation protection, and so checking on receipt or on packed finished product. We carry both technologies, and we do not offer them for the same question. What TERAHERTZ does well.
Thickness and dissolution are two attributes
The link exists and it is documented. A study co-signed by three Food and Drug Administration researchers shows that when the process is driven on the trajectory measured in line — thickness and roughness — the dissolution profiles of deliberately upset batches match those of batches run under optimal conditions. But it covers one product, at laboratory scale. Transposing it to your formulation means repeating the demonstration and documenting it in the control strategy.
Sacher, Fink, Herndler, Stumptner, Peter, Zettl, Korang-Yeboah, Feng, Wu, Khinast, Journal of Pharmaceutical Innovation 18, 2023, pp. 1870-1878, doi:10.1007/s12247-023-09750-5. Three of the authors are affiliated with the FDA (Silver Spring), and the work was funded by its Office of Pharmaceutical Quality.
A rate quoted with its context is the one to work from
Figures circulating for this technology vary by a factor of ten, because they name different things: image acquisition rate, rate of usable measurements, number of tablets actually characterised. The only marker we quote comes from a peer-reviewed study: on a pan run followed for more than three hours, 1,128 OCT detections over a population of about 840 tablets. The industrial-scale throughput figures we could find come from a trade publication rather than a peer-reviewed article, so we do not carry them over. The useful question is in any case how many tablets per hour, in my pan, with what analysis behind it.
Lin, Dong, Markl, Williams, Zheng, Shen, Zeitler, Journal of Pharmaceutical Sciences 106(4), 2017, doi:10.1016/j.xphs.2016.12.011.
What it changes, in practice
The gain shows in four places, and a site rarely combines them all: knowing which one concerns you is the purpose of the scoping.
Cycle time, when coating is run with a safety margin on weight gain. Stopping on a uniformity criterion rather than on a sprayed quantity removes that margin on the batches that do not need it. Consumption of coating suspension, which follows directly. The rework and reject rate, when insufficient uniformity is only discovered at final testing. And the strength of the dossier. A trajectory recorded on thousands of tablets is evidence of a different nature from a check on ten units.
The calculation is made on your own history, durations, weight gains and rework rates, during the scoping. It sometimes concludes that an already well-centred process has little to gain, and that is a useful answer.
Frequently asked questions
Does the instrument have to be calibrated on our tablets?
In the chemometric sense, no: there is no model to build on a representative population. The refractive index of the coating is needed to convert the optical path into a physical thickness, and the measurement is then verified on tablets whose thickness has been established by section. That is a verification rather than a calibration, and the difference shows in effort as in maintenance.
Does it work on a coloured coating?
On a lightly tinted coating, generally yes. On a coating heavily loaded with titanium dioxide or iron oxides, it calls for a feasibility study on your own formulation, since no pigment threshold is published to date. It is the first point to establish, ahead of the pan itself.
Can dissolution be predicted from thickness?
For a given product in an established domain, yes, and NIR also predicts it on an intact tablet: predicting dissolution. The demonstration exists in the literature, co-signed by three FDA researchers and funded by its Office of Pharmaceutical Quality — but at laboratory scale. Transposing it means repeating the correlation on your formulation, on batches covering the real variability, and writing it into your control strategy. It is a method validation project with an ambitious objective and a marked out path. Validating an analytical procedure.
Can it be installed on an existing pan?
Often yes, and it is the first point to establish, since it changes the cost entirely. What decides: the existence of an optical access, the possibility of fixing a measurement head rigidly, and the eccentricity of the drum. A new penetration on qualified equipment brings a requalification, costed at scoping.
And for an enteric or modified release coating?
That is exactly where the stake is highest, since thickness there is a functional attribute rather than a cosmetic one. It is also the case where the value of an in-process uniformity measurement defends itself most easily inside a company, because the regulatory argument meets the economic one.
What happens if the measurement drifts during a campaign?
Here, a drift is almost always mechanical or optical, working distance, a fouled window, a loosened mount, rather than a model drift. The diagnosis is simpler for it, and the correction faster. It calls for having planned it, with a periodic check on a standard and monitoring of the working distance.
Tell us the composition of your coating and the diameter of your pan. That is where it starts.
Forty-five minutes is enough to know whether the coating is measurable on your product, what the installation would call for, and what the gain would represent on your cycle times. Where the pigment load calls for another route, we will say which.