Granulation Loss on Drying (LOD): In-Process Moisture Control

Granulation Loss on Drying is the percentage of a granulation sample's initial mass lost during a specified drying procedure, which can include water and other volatile substances. It is a gravimetric result: the test measures a change in mass, not the chemical identity of everything removed.

To calculate LOD, subtract the final net sample mass from the initial net sample mass, divide the difference by the initial net sample mass, and express the ratio as a percentage. Net sample mass means gross mass minus the vessel tare. Using consistent mass units and a stable tare keeps the calculation focused on the sample; see tare-verified weighing. This initial-mass basis differs from a dry-mass basis, which uses the final dried sample mass as the denominator. The two bases are not interchangeable.

In manufacturing practice, LOD commonly supports decisions during granulation drying and checks after subsequent holding or handling. A result belongs to the particular sample, the process stage at which it was taken, and the drying method used. It is not an intrinsic, method-independent moisture value that can be transferred unchanged between sampling points or test procedures.

Sample drying loss is also distinct from manufacturing yield loss. Granules retained in equipment or lost during transfer affect material recovery, not the analytical LOD calculation. A drying mass balance addresses the broader accounting of material through the drying operation.

Chemically, Karl Fischer determines water specifically, whereas LOD can include other volatiles. The LOD versus Karl Fischer comparison explains why the results answer different questions.

In granulation development, residual moisture is commonly treated as a potential critical material attribute because it can influence flow, compaction, sticking and subsequent product stability. The direction and magnitude depend on the formulation and manufacturing conditions. Moisture may support deformation and particle bonding during compression, yet also promote cohesion or adhesion to tooling. Further drying may reduce sticking in one formulation but produce more friable granules or weaker compacts in another. Moisture-sensitive ingredients may also deteriorate during storage. Lower LOD is therefore not inherently better.

The practical reasoning runs from drying conditions to granule condition, then to downstream performance. Changes in drying exposure can alter moisture distribution and granule behaviour; those changes may affect feeding, die filling and compression. A useful endpoint connects a representative measurement with product- and process-specific evidence about those outcomes. The LOD result is an input to the decision to stop, continue or hold drying—not the decision itself. Development evidence commonly links endpoint results to downstream trials and stability observations rather than assuming that a passing value predicts every quality attribute.

Acceptable LOD also leaves other contributors unresolved. Particle-size distribution can influence flow and packing; binder distribution can affect granule strength and bonding; lubrication can alter compaction and ejection. These attributes help explain why batches with similar LOD may perform differently. Moisture control supports process understanding, but does not replace it.

Choose the measurement by the question it answers, not by the instrument’s moisture label. The comparison below describes analytical practice: oven LOD and halogen or infrared gravimetric analyzers implement drying-loss measurement; the other approaches measure or estimate different properties.

Oven and gravimetric moisture-analyzer results depend on the applied drying conditions, sample preparation and endpoint rule. Matching displayed units does not establish method equivalence. A method comparison therefore considers the complete procedures and their performance on the intended granules, rather than treating an instrument change as a simple substitution.

For LOD versus Karl Fischer, the central distinction is mass lost under a drying procedure versus water determined chemically. Other volatile substances can evaporate, and heating can induce sample changes that affect net mass loss. Consequently, drying loss cannot automatically be interpreted as water alone. Karl Fischer’s water selectivity does not remove the need for suitable sample preparation and assessment of matrix interference.

Water activity addresses availability rather than drying loss. Near-infrared results are model-based estimates calibrated against a reference method; the PAT section below explains how that calibration relationship is reviewed and maintained.

Binding US drug requirements

For US human drugs under 21 CFR 211, 21 CFR 211.110(a) requires written procedures to be established and followed describing in-process controls and tests or examinations on appropriate samples of in-process materials from each batch. This is the binding procedural framework for the US drug discussion, not a reason to select LOD without considering its usefulness for the material and process.

For US human drugs under 21 CFR 211, 21 CFR 211.110(c) assigns approval or rejection of in-process materials during production to the quality control unit.

Practice and implementation choices

Method selection, drying-endpoint development and suggested digital workflows are practical control-design choices in this discussion, not agency guidance. Treat LOD as a candidate within the control strategy: its usefulness depends on the process and material. A familiar test name is not, by itself, a justification for using its result to decide whether drying should continue or stop.

Software support and human decisions

As a vendor-neutral implementation practice, keep numerical result evaluation separate from authorized human review and disposition. A result classified as within range is an input to that review, not the decision itself. Design an electronic batch record workflow to support records and routing while leaving approval and release with authorized people. Describe the calculated status and the disposition status separately so that an operator does not mistake a completed calculation for permission to move material forward. These workflow choices do not replace applicable legal obligations and do not by themselves establish site compliance.

For US human drugs under 21 CFR 211, 21 CFR 211.160(b) requires laboratory controls to include scientifically sound and appropriate specifications, standards, sampling plans and test procedures. For US human drugs under 21 CFR 211, 21 CFR 211.110(b) requires valid in-process specifications to be consistent with final drug product specifications and derived from previous acceptable process averages and variability estimates where possible.

Build the product-specific justification

In industry practice, an LOD range is justified against the formulation, drying technology, sampling location, method conditions, measurement variability and downstream performance. Evidence from development and process validation can connect the measured result with relevant granule and finished-product attributes. The justification should explain both the process relationship and how reliably the test distinguishes acceptable from unacceptable conditions.

Copying another product’s range overlooks differences in materials, processing and measurement. Historical passing-batch minima and maxima describe what happened; alone, they do not establish acceptable boundaries. They can reflect limited operating experience or sampling and analytical variation rather than the limits of acceptable product performance.

Separate the decisions

Control strategies commonly distinguish these roles; local procedures identify the decision owners rather than treating every boundary as a pass/fail limit.

There is no assumed numerical margin between these controls. Their separation depends on process behavior, measurement capability and the consequences of each decision.

Make the proposal reviewable

A practical review package links development evidence, method identity and revision, sampling rationale, variability assessment and the proposed decision rule. Explain which result is compared with which criterion, who reviews it and what action follows. Through change control, a changed formulation, drying process or analytical method commonly triggers reassessment of comparability. The previous relationship between LOD and performance is a hypothesis to reassess, not an assumption to carry forward unchanged.

Collect a sample that represents the process

For US human drugs under 21 CFR 211, 21 CFR 211.160(b)(2) requires representative and properly identified in-process samples.

A practical sampling plan identifies the process stage, relevant dryer locations, sampling tool, container and transfer conditions. Select locations to address expected spatial variation rather than convenience alone. A sample pull prompt can turn these instructions into a repeatable collection checklist.

During handling, granules can lose moisture to, or absorb moisture from, their surroundings. Limit unnecessary exposure and follow the method’s container and transfer instructions. Record sampling and testing times through the sample login handoff, so elapsed time is visible. Establish handling allowances for the material and method rather than assuming a universal hold-time limit.

Execute the defined test

For US human drugs under 21 CFR 211, 21 CFR 211.194(a)(2) requires testing-method suitability to be verified under actual conditions of use.

In practical LOD execution, identify the method version and follow its sample preparation, loading, spreading, tare handling, drying program and endpoint rule. Keep these details method-specific: a different sample layer or endpoint interpretation can change the result even when the granules have not changed.

For US human drugs under 21 CFR 211, 21 CFR 211.160(b)(4) prohibits using instruments, apparatus, gauges and recording devices that do not meet established specifications. As an execution check, confirm instrument status before loading the sample.

Make the result reviewable

For US human drugs under 21 CFR 211, 21 CFR 211.160(a) requires the laboratory-control requirements in that subpart to be followed and documented at the time of performance.

Retain enough run information to distinguish a changed material from a changed test: sample identity, handling times, instrument identity, method version, masses, actual settings, endpoint information and execution observations. Make interruptions or departures visible to the reviewer rather than reducing the run to its final percentage.

Practical workflow—not a prescribed regulatory sequence: Define sampling → Collect and identify → Protect and transfer → Confirm instrument and method → Run LOD → Document result and exceptions → Human review → Process decision

Calculation case: vessel-plus-sample readings

Assume a pharmaceutical granulation test uses a stable vessel tare, no sample is spilled, and all masses are expressed in grams. Let T be the empty vessel tare, I the initial vessel-plus-sample mass, and F the final vessel-plus-dried-sample mass. These are symbolic readings, not production measurements. The net-versus-gross reconciliation is:

• Initial net sample mass = I − T.
• Dried net sample mass = F − T.
• Mass lost = (I − T) − (F − T) = I − F.
• LOD (%) = [(I − F) / (I − T)] × 100.

This applies the initial-mass percentage relationship introduced above. Dividing instead by F − T reports loss relative to the dried residue: a different reporting basis. For positive mass loss and a positive dried mass, that denominator is smaller and the resulting percentage is larger.

Correct arithmetic alone establishes neither batch representativeness nor water specificity. Those interpretation boundaries depend on the earlier definition and sampling controls, not on cancellation of the tare.

Sampling case: different dryer locations

Assume food or supplement granules share the same formulation, process stage, handling conditions and test method. Results are reported as percentage mass loss. A sample near one dryer location gives a lower result; a sample elsewhere gives a higher result. No acceptance limit is assumed.

An aggregate result may conceal this spatial variation rather than demonstrate a uniform endpoint. To assess coverage, examine the sampling map, sampled depths, collection sequence and timing, and the rationale for how much material each location represents. Link each result to its location through the sample login handoff, so the comparison remains interpretable.

A practical response is to preserve location-specific results, assess the sampling rationale, and route the endpoint decision through the site's defined product-specific process. The difference alone is not an instruction to continue drying; this operational example does not establish a US drug obligation for food or supplement manufacture.

Process analytical technology (PAT) can use near-infrared spectra and a chemometric model calibrated against a named reference method—for example, the site's gravimetric granulation LOD procedure. A model trained against LOD predicts that reference response, not automatically water-specific content. Keep the prediction distinguishable from a directly measured LOD result.

Practical model review commonly examines whether the calibration population represents the formulations, moisture range and physical granule conditions encountered in use. Changes to the reference drying method can change the response being predicted. Comparing subsequent paired predictions and reference results helps detect model drift; unfamiliar material conditions warrant review rather than unquestioned extrapolation.

Binding laboratory-record requirements

• For US human drugs under 21 CFR 211, 211.194(a)(1) requires laboratory sample records identifying source, quantity, lot or other distinctive code, sampling date and receipt date.
• For US human drugs under 21 CFR 211, 211.194(a)(4) requires a complete record of test data, including graphs, charts and spectra from laboratory instrumentation.
• For US human drugs under 21 CFR 211, 211.194(a)(5) requires records of test calculations, including units, conversion factors and equivalency factors.
• For US human drugs under 21 CFR 211, 211.194(a)(8) requires a second person's initials or signature showing review of original records for accuracy, completeness and compliance with established standards.

Example evidence relationships

This implementation diagram illustrates data integrity by design, not a legally prescribed database schema. Production context tagging connects measurements to material and process state.

LotID → SampleID [location, quantity, sampledAt, receivedAt] SampleID → DirectTestID [MethodVersion, rawMasses, calculation, LODResult] SampleID ↔ SpectrumID → PredictionID [ModelVersion, predictedLOD] ModelVersion → CalibrationSetID → paired SpectrumID + DirectTestID DirectTestID / PredictionID → ReviewID [reviewer, decision]

Separate result identifiers preserve the distinction between measurement and prediction while allowing reviewers to follow each result back to its supporting records.

A proposed connection could send instrument results to a laboratory information management system (LIMS), then pass reviewed results and status to a manufacturing execution system (MES). For this MES–LIMS integration, reconcile sample and batch identifiers, method versions, units, reporting basis and raw-data references before using an imported result. These are vendor-neutral design considerations, not claims about existing software capabilities.

Worked recovery: transmission interrupted

Assume the instrument completes an LOD test, but transmission stops before acknowledgment. The result is percentage drying loss; its underlying masses are in grams. A practical recovery sequence is:

1. Confirm the original completed run and preserve its result, masses and raw-data reference.
2. Check whether LIMS already holds the receiving record; missing acknowledgment does not establish missing data.
3. Reconcile the run, sample and batch identifiers, method version, units and reporting basis against that record or the pending message.
4. If retransmission is needed, retain the original run identity and reconcile the replay so it cannot create a duplicate result or trigger a second MES decision.

Resending transfers the existing result; repeating the physical test creates a new analytical event. Record which action occurred.

For US human drugs under 21 CFR 211, 21 CFR 211.160(a) requires deviations from written specifications, standards, sampling plans, test procedures or other laboratory-control mechanisms to be recorded and justified.

Keep unresolved test or interface exceptions distinct from established material rejection in exception handling. The human decision boundary described above still applies: an alarm, failed import or calculated flag is not itself rejection or release.

For US human drugs under 21 CFR 211, 21 CFR 211.110(d) requires rejected in-process materials to be identified and controlled under a quarantine system preventing unsuitable manufacturing or processing use.

In routine gravimetric troubleshooting, an unexpected LOD result suggests an investigation path—not a proven cause. Use targeted checks before changing drying conditions or interpreting the result as a process shift.

Good analytical practice preserves the original result and uses a reasoned investigation rather than repeating tests until an acceptable value appears or silently replacing an inconvenient result. Follow the site's out-of-specification handling process where applicable; document the rationale for further testing.

Use this practical checklist before relying on LOD for a process decision. For explanations, refer back to regulatory scope, specification setting, execution, and exception recovery.

• Scope: Are the product category, jurisdiction and intended process decision identified?
• Measurand: Is the intended measurement clearly stated, without treating drying loss as water-specific?
• Sampling: Does the rationale address location, timing, representativeness and sample handling?
• Method: Are the method version, instrument configuration and evidence of suitability identifiable?
• Specification: Is the acceptance basis justified, with process-adjustment signals distinguished from acceptance criteria?
• Original data: Can reviewers access source records and their links to the relevant batch record?
• Calculations: Have tare treatment, units, calculation basis and rounding been reviewed?
• Interfaces: Are transferred results reconciled with instrument records, including failed, duplicate or corrected transfers?
• Ownership: Are the people authorized to interpret results, resolve exceptions and approve the next action named?

Readiness test: Can a reviewer reconstruct which material was sampled, how it was tested, what happened during exceptions and why the next process decision was made? Use linked records and audit trail review to answer. Checklist completion is not compliance certification; approval and release remain human responsibilities.