Granulation Loss on Drying (LOD)

Loss on drying is the percentage weight loss when a sample is heated to a defined temperature for a defined time. The compendial reference is USP <731>; Karl Fischer (USP <921>) is the more specific water-only measurement. NIR moisture is increasingly used as in-line or at-line PAT for continuous control of end-of-drying. The end-of-drying LOD is the gate to release granules from the dryer — fail and the bed is re-dried, pass and the granules move to mill, blend and press.

LOD is not the same as water content. LOD captures everything volatile at the test temperature — water plus residual organic solvents, plus any volatile excipients (e.g. menthol in some chewables). Karl Fischer measures water specifically through a redox reaction with iodine, so KF reads lower than LOD when organic solvents remain. In a wet-granulation line using water as the only solvent the two converge; in an organic-solvent granulation (rare in solid dose, common in some film coatings) the difference can be significant.

• Typical solid-dose LOD spec: 1.0–3.0% depending on product and downstream tolerance.
• LOD includes any volatile, not just water; KF measures water only.
• NIR moisture validated under ICH Q2/Q14 can replace destructive sampling at full PPQ.
• Water activity (USP <1112>) supplements LOD for microbial-risk decisions on hygroscopic products.
• LOD failures are one of the most common in-process reasons for batch rework in solid-dose lines.

Moisture in granules has multiple downstream effects, each of which can cause a release failure or a stability OOS. Too high: granules cap or laminate on the press, tablet hardness varies, dissolution shifts, microbial limits may fail at long shelf life. Too low: granules embrittle, fines proliferate, content uniformity degrades, and some actives (e.g. amorphous APIs requiring a hydration shell) destabilise. The spec is therefore a band, not a maximum.

Most commercial solid-dose lines run IR-balance LOD as the at-line release method, bridged to oven LOD or Karl Fischer during validation. NIR is layered on top as a continuous indicator that triggers the official LOD sample at the right moment. The choice is documented in the validated control strategy and any switch (IR to oven, oven to NIR-primary) is a change control with a comparability study.

An LOD spec must be justified by data, not copied from a similar product. Development should map LOD against dissolution, hardness, friability and stability across three drying time points (under-, target- and over-dried lots) at two manufacturing scales. The acceptance band is the LOD range that delivers all CQAs at release and through shelf life with statistical margin. ICH Q6A requires the spec be tightened where data supports it; the FDA Question-Based Review reviews this justification explicitly for ANDA filings.

Stability data is the long pole. Hydrolysis-prone actives (esters, amides, certain peptides) push the LOD upper limit down because residual moisture accelerates degradation in the solid state. Conversely, an active that depends on a hydration-shell for chemical stability needs a lower limit too. The dossier shows the stability assay vs RM at 6, 12, 24 and 36 months from multiple commercial-scale batches.

• Sample at end-of-drying from multiple bed locations (stratified) — top, middle and base of the bed.
• Use the validated method — don't switch IR for oven mid-campaign without a change control.
• Bridge NIR to KF during PPQ for the regulatory record, then reduce to NIR-primary in CPV.
• Tie LOD spec to dissolution and stability data — not picked from thin air.
• Record LOD in the batch record as a CQA, not just an in-process flag.
• Track LOD test elapsed time — long delays between sample and result allow moisture re-equilibration with room air, biasing the number.
• Calibrate IR-balance against the reference method on a defined frequency; check daily with a moisture standard.

• Sampling only the top of the bed — bottom may still be wet, especially in fluid-bed dryers with poor distributor design.
• Using IR for batch release without bridge to compendial method — fine in development, regulatory finding at inspection.
• Specs copied from a similar product without dissolution justification — risks both OOS and over-conservative wasted drying time.
• No NIR/PAT trending — drying time creep across a campaign goes unnoticed.
• Treating LOD pass as moisture pass — volatile residual solvents missed entirely.
• Test temperature too low — water of crystallisation not removed, false-low result.
• Sample bottles uncapped on the bench — atmospheric moisture re-equilibrates and biases the result.
• Single LOD per bed — never enough to characterise the bed; OOS investigations show wide intra-bed variation.

• Solid-dose pharma — LOD spec is a standard release attribute on every wet- or dry-granulated tablet and capsule product.
• Nutraceutical chewables — moisture drives shelf-life and microbial limit; spec set by water-activity correlation.
• Effervescent — extremely low LOD (<0.5%) needed; even small moisture excursions trigger early reaction.
• Veterinary medicated articles — LOD ties to active stability in feed environments with variable storage humidity.
• Food / instant beverage powders — moisture drives flow, caking, microbial limits and reconstitution time.
• Battery cathode powders (non-pharma) — analogous LOD method protects against electrolyte side reactions.

NIR moisture is the dominant PAT tool for granulation drying. A probe in the dryer collects spectra at 1–10 s intervals; a validated chemometric model converts spectra to %moisture. The model is built on a calibration set spanning the full anticipated moisture range, validated under ICH Q2(R2) and the FDA PAT framework. Maintenance is real work: instrument drift, lamp ageing, probe fouling and matrix shifts all require periodic re-validation. A NIR programme without a documented model lifecycle is an audit finding waiting to happen.

Data integrity controls follow ALCOA+. The LOD result must be attributable to the analyst and instrument, contemporaneously recorded, original (not transcribed from a paper note), accurate (within method tolerance), complete (all integrations and decisions captured), enduring (retained for the statutory period), and available for audit. Modern MES integrations stream the result from the IR balance or NIR system directly to the batch record, removing the transcription step entirely — and removing the opportunity for re-test-until-pass behaviour that has driven multiple recent warning letters.