Quantitative characterization of lactose crystalline forms

Importance of the topic

The crystalline form of lactose in raw materials and excipients strongly influences manufacturability, stability, and performance of pharmaceutical and food products. Rapid, robust methods that quantify different lactose forms (alpha-lactose monohydrate, alpha-lactose anhydrous, and amorphous lactose) are therefore valuable for incoming material control and in-process monitoring. Fourier transform near-infrared spectroscopy (FT-NIR) offers a fast, non-destructive and operator-insensitive approach for routine characterization of these physical forms.

Objectives and overview of the study

This application note demonstrates the quantitative use of FT-NIR to distinguish and quantify the three common forms of lactose in binary and co-varying ternary mixtures. The aims were to (i) establish spectral differences among forms, (ii) build and validate chemometric calibration models (PLS) for binary and ternary mixtures, (iii) determine limits of quantification, and (iv) assess applicability of FT-NIR calibrations relative to reference techniques such as X-ray powder diffraction (XRPD).

Methodology

Sample preparation:

• Starting materials: alpha-lactose monohydrate and alpha-lactose anhydrous (commercial sources) and spray-dried amorphous lactose.
• Amorphous lactose production: 10% w/w aqueous solution of alpha-lactose monohydrate spray-dried (inlet 187 °C, outlet ~90 °C, airflow 0.34 m3/min), then vacuum-dried at 50 °C for 30 min and stored over calcium sulfate.
• Sieving: crystalline forms sieved to 38–125 µm; amorphous material screened to exclude particles >150 µm to limit agglomeration.
• Mixtures: binary mixtures (three pairwise combinations) prepared and independently validated; ternary mixtures prepared following a triangular design, avoiding >40% amorphous content to minimize agglomeration.

FT-NIR data collection:

• Instrument: Thermo Scientific Antaris FT-NIR Solid Sampling system (integrating sphere, bottom-read vials).
• Measurement mode: diffuse reflectance.
• Spectral range: 4000–12000 cm-1 (specific calibration subregions used per model).
• Resolution: 4 cm-1; co-averaged scans: 90; typical collection time ~67 s; detector: InGaAs.

Chemometrics and spectral pretreatment:

• Software: Thermo Scientific TQ Analyst; algorithm: PLS-I (Partial Least Squares).
• Model optimization: Region Select to pick spectral windows and PRESS plots to select number of PLS factors.
• Pretreatments: Multiplicative Scatter Correction (MSC), Norris derivatives, and Savitzky–Golay second derivatives (varied filter lengths: e.g., third-order 17- or 25-point segments depending on model).

Instrumentation used

The study used a Thermo Scientific Antaris FT-NIR Solid Sampling system with integrating sphere and InGaAs detector. Measurements were configured as bottom-read reflectance through standard sample vials. The authors note that Antaris II is a newer model offering improved speed and performance compared to the instrument used in this study.

Results and discussion

Spectral discrimination:

• Full-range and second-derivative FT-NIR spectra show clear, reproducible differences among alpha-lactose monohydrate, alpha-lactose anhydrous, and amorphous lactose, enabling quantitative modeling.

Binary mixture calibrations (representative statistics):

• Amorphous + alpha-lactose monohydrate (model for alpha-lactose monohydrate): R2 = 0.9992; RMSEC = 1.29; RMSEP = 1.98.
• Amorphous + alpha-lactose anhydrous (model for alpha-lactose anhydrous): R2 = 0.9995; RMSEC = 1.05; RMSEP = 2.12.
• Alpha-lactose anhydrous + alpha-lactose monohydrate (model for alpha-lactose anhydrous): R2 = 0.99999; RMSEC = 0.136; RMSEP = 0.718.

Ternary mixture calibrations:

• Alpha-lactose monohydrate: 5-factor PLS, region 5249–8916 cm-1; R2 = 0.9981; RMSEC = 1.63; RMSEP = 3.95.
• Alpha-lactose anhydrous: 5-factor PLS, region 4902–5768 cm-1; R2 = 0.9971; RMSEC = 2.04; RMSEP = 3.40.
• Amorphous lactose: 3-factor PLS, same region as crystalline models; R2 = 0.9817; RMSEC = 2.26; RMSEP = 2.02.

Limits of quantification and performance notes:

• LOQs for components in ternary matrices were reported in the range ~4.1% to 6.4% (estimated as 3×SD from low-level replicate measurements).
• Models for crystalline forms (monohydrate and anhydrous) were generally stronger and more robust than models for amorphous lactose because crystalline spectral responses were more distinct and intense.
• Application to real, homogeneous systems was supported by an earlier comparison: NIR calibrations built from synthetic mixtures applied to milled sucrose samples produced trends consistent with XRPD, though a bias relative to XRPD was observed and is correctable.

Benefits and practical applications of the method

• Speed and throughput: FT-NIR spectra are collected rapidly (under a minute per sample in this study) and require minimal sample preparation.
• Non-destructive and reagent-free analysis suitable for raw material screening and at-line checks on production floors.
• Operator-independence: instrument + chemometric models enable routine use by non-expert personnel.
• Ability to quantify multiple solid-state forms simultaneously (co-varying ternary matrices), making FT-NIR useful for incoming goods control, stability monitoring, and troubleshooting manufacturing variation linked to polymorphism or amorphicity.

Future trends and possibilities for use

• Instrumentation advances: faster detectors and improved FT-NIR models (e.g., Antaris II) will increase throughput and sensitivity.
• Model transferability and robustness: development of large, representative calibration sets built from real production lots will enhance applicability across sources and lots; standardization procedures (transfer or domain adaptation) will facilitate deployment across instruments/sites.
• Integration with process analytics: at-line, on-line, or in-line FT-NIR combined with process data and multivariate control can enable real-time release testing and tighter process control.
• Advanced chemometrics and machine learning: regularization, non-linear methods, and domain adaptation can improve amorphous content prediction and reduce bias versus reference methods.
• Hybrid approaches: combining FT-NIR with complementary techniques (XRPD, DSC) for calibration anchoring and orthogonal confirmation will strengthen confidence in measurements for regulatory contexts.

Conclusions

FT-NIR is demonstrated as a practical, high-performance technique to quantify alpha-lactose monohydrate, alpha-lactose anhydrous, and amorphous lactose in binary and ternary mixtures. Calibrations based on PLS models with standard scatter- and derivative pretreatments yielded high correlation coefficients and acceptable prediction errors for crystalline forms, with somewhat lower robustness for amorphous content. Limits of quantification in the ternary matrices were on the order of a few percent. FT-NIR offers a rapid, non-destructive, and production-suitable approach for raw material characterization and in-process control; complementary reference methods (e.g., XRPD) remain useful for method validation and bias correction.

Reference

1. Hirsch J., McCarthy W.J., Luner P.E., Patel A.D., Seyer J.J. Quantitative characterization of lactose crystalline forms. Thermo Fisher Scientific application note (AN50775_E). Thermo Fisher Scientific, Madison, WI, USA.