Decide what the measurement must justify
A batch shows stable chamber pressure, and the instrumented vials have warmed toward the shelf. The operator asks whether primary drying is complete. Neither observation alone establishes that the last ice-containing vial has finished. The monitoring strategy must define what conclusion each signal supports, which product population it represents and how disagreement will be handled.
Start with the decisions: controlling pressure, protecting product temperature, identifying a phase transition, documenting cycle execution or supporting a release-related conclusion. Different decisions may require different evidence. A control sensor can be suitable for regulating the equipment without demonstrating product uniformity across the batch.
The objective is a coherent combination of measurements and justified interpretations. Product temperature, shelf temperature, pressure and analytical estimates should reinforce a process explanation rather than compete as interchangeable “truths.” Their limitations belong in development, qualification and routine instructions, where they can influence decisions before a difficult batch occurs.
Define the measurement chain
For every important signal, identify the measured quantity, sensor principle, location, operating range, calibration basis and data path. Distinguish the sensor output from corrected values, displayed values and calculations derived from several inputs. A trace labelled “product temperature” may be a direct measurement, external-vial measurement or model estimate.
Calibration should address the range and uncertainty relevant to the decision. Laboratory accuracy alone does not establish performance after installation, repeated sterilization, exposure to vacuum or connection to a different acquisition system. Include response time, drift, mounting effects and potential changes following maintenance.
Maintain traceability from the raw observation to the batch record. Sampling interval, filtering, averaging and time synchronization can alter the apparent duration of an excursion or endpoint transition. A smooth trend is not necessarily a more informative trend. Select processing that preserves the information required for control and investigation, and record how derived results were produced.
Interpret shelf measurements within their limits
Shelf sensors describe the thermal system at their installed positions. Depending on design, the control measurement may be associated with heat-transfer fluid rather than the product-contact surface. Mapping can establish distribution and dynamic behaviour under specified conditions, but these are equipment characteristics, not a direct map of product temperature.
Consider heating and cooling transitions as well as holds. Fluid distribution, shelf construction, load and refrigeration demand can influence how the surface responds. If only a setpoint is retained in the batch report, an important distinction between commanded and achieved performance is lost.
Use shelf information as context for product and pressure observations. An unusual product-temperature rise may follow a genuine thermal excursion, a changed heat-transfer environment or sensor movement. Comparing actual fluid, shelf, product and phase records helps separate these possibilities. Avoid the ambiguous expression “freeze-drying temperature” when the decision requires a specific location and physical meaning.
Place product probes to answer physical questions
Thermocouples and other product-contact sensors can provide valuable development data, but an instrumented vial may behave differently from an uninstrumented one. Probe insertion can influence nucleation, local geometry and heat transfer. A single sensing junction also measures its immediate environment rather than the entire fill or sublimation interface.
Select positions using the expected sources of variability. Peripheral vials, central populations, shelf differences and partial loads may create different concerns. The hottest population is not automatically the last to complete primary drying. A probe arrangement intended to investigate maximum temperature may therefore be insufficient for endpoint assessment.
Document placement and retention through loading, freezing, drying and stoppering where applicable. A sensor that moves as the cake forms can generate a believable but misleading trace. Establish how suspect channels are identified, retained in the record and excluded from a particular interpretation. Do not simply remove inconvenient signals without a documented technical basis.
Evaluate wireless and non-contact options carefully
Wireless probes may reduce cabling or support positions that are difficult to instrument, but “wireless” does not mean “non-invasive.” A wireless device placed inside a vial can still perturb nucleation or product geometry. External-vial sensors avoid some direct product contact but measure a different physical location.
A model may translate external measurements into an estimate of internal temperature. That estimate requires verification for the relevant vial, fill, formulation and operating conditions. Transmission reliability, battery or power behaviour, clock alignment and data recovery must also be considered where those features apply.
Choose the technology by the decision it improves, not by novelty. Assess compatibility with loading equipment, sterilization, cleaning and the aseptic boundary. Published demonstrations show useful possibilities, but do not qualify a system for every commercial application. Retain a clear distinction between research performance, installed-system qualification and an approved use in routine process control.
Understand capacitance and Pirani readings
A capacitance manometer determines pressure through diaphragm deflection and is substantially independent of gas species. A Pirani gauge responds to heat transfer through the gas, so its indication depends on gas composition as well as pressure. During freeze-drying, changing proportions of water vapour and noncondensable gas can therefore change the relationship between the instruments.
This difference is useful when deliberately interpreted. Comparative pressure monitoring can indicate declining water-vapour contribution near the end of primary drying. A stable absolute pressure reading and a changing Pirani signal are not necessarily contradictory; they can reflect pressure control maintaining total pressure while gas composition evolves.
Avoid using a universal ratio, fixed gap or convergence time without supporting studies. Calibration, installation, sensor temperature, gas admission and equipment conditions can influence the comparison. Nor is total chamber pressure automatically the water-vapour partial pressure used in a mass-transfer model. Define the approximation and its limits when those quantities are treated as similar.
Distinguish a pressure indicator from a product endpoint
An endpoint is an operational conclusion supported by evidence, not merely a recognizable bend in a trend. Specify whether the decision concerns completion of ice sublimation, suitability for the secondary-drying ramp or achievement of final product moisture. These are different questions and may occur at different times.
Develop the primary-drying decision against relevant product observations. Sampling studies, independent moisture measurements, temperature populations and comparative pressure behaviour can establish what the signal means under the intended conditions. Include the slower portion of the load and realistic sources of variability. The instrumented vial that finishes first should not define completion for the batch.
If the control strategy uses a persistence period or additional hold, justify it from development and transfer evidence. Specify what happens when channels disagree, a sensor fails or the expected transition is absent. An unexplained timer added to an uncertain signal can conceal the problem without demonstrating control.
Use MTM as an estimation method
Manometric temperature measurement, or MTM, interprets the chamber-pressure rise during a brief isolation from the condenser. A model can estimate quantities such as an effective product temperature and dried-layer resistance from the transient. It is a process analytical approach, not direct thermometry of every vial.
The result depends on the model, load, temperature distribution and quality of the pressure response. Small or heterogeneous loads can challenge the assumptions. Different vial populations can dominate different aspects of the fitted curve. Leakage, outgassing, valve performance and the thermal effect of isolation also require consideration when interpreting the transient.
Define applicability before using MTM for a decision. Examine fit quality, parameter plausibility and consistency with independent measurements. A software-generated value with many decimal places does not establish accuracy. Likewise, published vial counts and isolation times describe particular experiments; they are not universal instructions for another chamber, product or production configuration.
Select complementary analytical evidence
Additional approaches may evaluate water vapour, mass flow, heat flux, product characteristics or model-based state estimates. Their value depends on whether they provide information that reduces uncertainty in a specific decision. Adding more instruments can create more disagreement if their measurement principles and spatial coverage remain poorly understood.
For secondary drying, residual-moisture development generally requires an appropriate analytical method and sampling strategy. A primary-drying pressure transition does not by itself establish the final moisture range. Moisture, water activity and product stability are related but distinct concepts; do not substitute one measurement for another without a justified relationship.
The following original comparison is a planning aid.
| Evidence | Useful contribution | What it does not establish alone |
|---|---|---|
| Shelf or fluid temperature | Equipment thermal execution | Product temperature in every vial |
| Product-contact probe | Local product response | Unperturbed whole-batch behaviour |
| Capacitance manometer | Total pressure at its location | Water-vapour partial pressure in all conditions |
| Pirani comparison | Changing gas composition and endpoint indication | Final residual moisture or absence of ice everywhere |
| MTM estimate | Model-based temperature/resistance information | Direct spatial measurement of all vials |
| Product moisture assay | Water content in sampled material | Complete aseptic assurance or all other CQAs |
Distinguish PAT from routine monitoring
Routine recording of temperatures and pressure is necessary process information, but a historian does not turn every recorded parameter into process analytical technology. PAT connects appropriate analytical measurements and understanding to assessment or control of relevant process and product states. The intended analytical claim should be explicit.
FDA PAT guidance supports a science-based framework; it does not impose a particular instrument on every freeze-drying process. A simple, well-characterized monitoring approach may be appropriate, while an advanced model may be justified where it resolves material uncertainty. Identify which conclusions are scientific, which are quality-risk-management choices and which are applicable regulatory obligations.
For closed-loop decisions, define the complete path from signal to action. Include data-quality checks, allowable operating bounds, model version, failure response and manual intervention authority. Verify that an analytical or communication failure cannot silently become permission to advance the cycle. A validated control function must remain understandable to operators and reviewers.
Investigate disagreement in a practical case
During a routine run, product probes warm while the Pirani comparison still indicates substantial water vapour. One tempting explanation is that the pressure sensor has failed. Another is that the monitored vials dried earlier than the slower population. Both require evidence; selecting the interpretation that finishes the batch fastest is not a control strategy.
Review probe positions, freezing histories, calibration status, absolute pressure, gas admission and phase timing. Compare the signal pattern with characterized batches and assess whether the current load lies within the established monitoring scope. Preserve the original data and document the basis for any continued hold, interruption or escalation.
After the batch, targeted studies can distinguish a location bias from instrumentation or equipment behaviour. The corrective action may involve sensor coverage, endpoint logic, operator instructions or a process change. Do not rewrite a threshold around one unexplained run. Changes affecting the validated strategy require assessment before they become routine.
The investigation should also consider whether loading or container changes altered the relationship between the measured population and the batch. A monitoring strategy established for one fill height or tray arrangement may lose representativeness after a seemingly minor change. Document this connection in change assessments so that the sensor plan evolves with the process it is intended to describe.
Make the monitoring package reviewable
Before routine use, confirm that the package identifies what is measured, inferred and decided. It should contain sensor locations, calibration and uncertainty information, data-processing rules, the basis of endpoint criteria, known limitations and responses to failure. Link it to the relevant equipment qualification and product-process evidence.
Challenge foreseeable faults during suitable testing: loss of a channel, stale values, clock discrepancies, implausible estimates or a phase transition attempted before its conditions are met. The appropriate scope follows risk and system design. There is no universal sensor count, sampling interval or mandatory list of advanced PAT devices for all products.
Finally, ensure that reviewers can reconstruct the batch without relying on a screenshot. Retain meaningful raw data and metadata, actual profiles, events, overrides and relevant audit-trail information through the applicable record system. Red flags include ambiguous signal names, an endpoint defined only by operator intuition and a final-moisture claim based solely on Pirani convergence. A useful monitoring strategy states both what its evidence proves and what remains outside that evidence.
Sources and scope
Sources checked on 26 September 2026. Apply requirements within their jurisdiction and scope. Scientific evidence and engineering recommendations do not establish universal cycle settings. Examples are illustrative. For licensed documents, only public scope and edition were verified; research access limitations are recorded in the source register.
- Determination of end point of primary drying in freeze-drying process control (2010) — Scientific principle.
- Evaluation of manometric temperature measurement, a process analytical technology tool for freeze-drying: Part I, product temperature measurement (2006) — Scientific principle.
- Evaluation of manometric temperature measurement, a process analytical technology tool for freeze-drying: part II measurement of dry-layer resistance (2006) — Scientific principle.
- A non-invasive multipoint product temperature measurement for pharmaceutical lyophilization (2022) — Scientific principle.
- Detection of Collapse and Crystallization of Saccharide, Protein, and Mannitol Formulations by Optical Fibers in Lyophilization (2018) — Scientific principle.
- PAT — A Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance — Final guidance; PDF September 2004 — Guidance.
- EU GMP Annex 11 — Computerised Systems — Revision 1, January 2011 — Regulatory requirement.
