Use of simple standards for model transfer in quantitative analysis of intact tablets by near-infrared spectroscopy.
Journal of Pharmacy and Pharmacology
Objectives To transfer spectra measured on a slave instrument to a model created on a master instrument using simple pure compounds or mean sample spectra as the transfer standard. Methods Reflectancenear-infrared (NIR) spectra of 116 Sterwin(R) 500 mg paracetamol intact tablets (29 batches, 76.5-91.6% m/m active, nominal diameter 12.9 mm) were measured on three different FOSS NIRSystems instruments. Reference values for the active were measured according to the British Pharmacopoeia 2008. Instruments: A, SCS Masterlab spectrometer; B, XDS spectrometer fitted with a rapid content analyser (RC) and C, 6500 monochromator fitted with a RCA. The following transfer standards were used; avicel PH101, benzoic acid, methylparaen, paracetamol, sucorse, mean sample spectrum (48 paracetamol tablets), mean paracetamol tablets (other brands) and six different pharmaceutical tablets. Two containeers, a 12 mm diameter vial and a 50 mm diameter cell, were used to measure the powders. Spectra of tablets measured on instruments B and C were corrected using the difference spectrum of the transfer standad measured on A fand B or C respectively. These corrected spectra were then used to predict the partacetamol content using a model developed on A. Results A three-factor partial least squares regression (PLSR) calibration model (1100-2498 nm) using SNV + first-derivative spectral and pre-treatment on instrument A gave an RMSEP of 0.57% m/m. Direct transfer of spectra from instrument B gave an RMSEP of 0.81% m/m. Corrected spectra gave slightly better results: for example, avicel PH101 in vial (0.95% m/m). Transers between different instrument types (A and C) were less successful: direct transfer (0.97% m/m), avicel PH101 in vial (0.75%), avicel PH101 in cell (1.60% m/m), paracetamol in vial (0.54% m/m), sucrose in vial (1.07% m/m) and sucrose in cell (1.03% m/m). However, correction using the man sample spectrum as best (0.52% m/m). Many factors influenced the success of the transfer. Model selection was important; for example, increasing the level of smoothing during spectral pre-treatment generally improved transfer. Matching the transfer standard dimensions to that of the paracetamol tablets was also an important factor. In general the transfer standard needs to closely math the sample both chemically and physically. Conclusions Transfer using simple pure compounds, while useful, was not to be preferred over mean sample spectrum correction.
|Title:||Use of simple standards for model transfer in quantitative analysis of intact tablets by near-infrared spectroscopy.|
|Additional information:||Original article available from Pharmaceutical Press.|
|UCL classification:||UCL > School of Life and Medical Sciences
UCL > School of Life and Medical Sciences > Faculty of Life Sciences
UCL > School of Life and Medical Sciences > Faculty of Life Sciences > UCL School of Pharmacy
Archive Staff Only