1 Verification of model development technique for NIR-based real-time 2 monitoring of ingredient concentration during blending 3 Hiroshi Nakagawa a, * Q1 , Manabu Kano b , Shinji Hasebe c , Takuya Miyano a , 4 Tomoyuki Watanabe a , Naoki Wakiyama a 5 a Formulation Technology Research Laboratories, Pharmaceutical Technology Division, Daiichi Sankyo Co., Ltd., Kanagawa, Japan 6 b Department of Systems Science, Kyoto University, Kyoto, Japan 7 c Department of Chemical Engineering, Kyoto University, Kyoto, Japan A R T I C L E I N F O Article history: Received 4 December 2013 Received in revised form 9 April 2014 Accepted 9 May 2014 Available online xxx Keywords: Near infrared (NIR) Locally weighted partial least squares (LW- PLS) Process analytical technology (PAT) Blending process Real-time monitoring A B S T R A C T There has been a considerable research on the process analytical technology (PAT) and real-time monitoring based on NIR, but the model development is still an important issue and persons in charge have difficulty in building good models. In this study, to realize efficient NIR-based real-time monitoring of ingredient concentration and establish a model development method, we investigated the effect of a calibration set, spectral preprocessing, wavelengths, and other factors on the prediction error through pilot and commercial scale blending experiments. The results confirmed that the small prediction error was realized by a calibration set, including dynamic measurement spectra acquired with the target blender. In addition, the results demonstrated that locally weighted partial least squares (LW-PLS) achieved the smaller prediction error than conventional PLS. The present study has also clarified that spectral preprocessing methods and wavelengths selected to build a model affect the prediction error of ingredient concentration interactively. A wide wavelength range should be selected when the spectral preprocessing does not lessen the effect of baseline variation, while a narrow wavelength range should be selected when it strongly decreases the effect. ã 2014 Published by Elsevier B.V. 8 1. Introduction 9 To ensure the content uniformity of drug products, the blend 10 uniformity in the blending process needs to have the concentration 11 of the active pharmaceutical ingredient (API) almost the same as the 12 target value at representative points in the blender (e.g., the API 13 concentration is not less than 90% at all points) and consistent (e.g., 14 the relative standard deviation is not more than 5.0%) (FDA, 2003). In 15 the past, manufacturing conditions to ensure blend uniformity were 16 investigatedthrough process development study, with three batches 17 being validated by using commercial scale equipment. Even if the 18 manufacturing conditions are well established in the validation, 19 unexpected issues may occur due to various disturbances such as 20 variations in the raw material properties and environments. For 21 example, humidity affects the blending time needed to achieve the 22 desired blend Q2 uniformity (El-Hagrasy et al., 2006). 23 To prevent such issues, it is useful to develop a real-time 24 monitoring technique to control the blend uniformity in routine 25 commercial manufacturing. Rapid measurement techniques, 26 including real-time monitoring, have been studied with enthusi- 27 asm as a part of process analytical technology (PAT) (FDA, 2004). 28 NIR spectrometers, which are popular PAT tools (Reich, 2005; 29 Roggo et al., 2007; Jamrógiewicz, 2012), have been applied to rapid 30 measurement and real-time monitoring of the blend uniformity of 31 various components (Nakagawa et al., 2013). The studies on rapid 32 measurement and real-time monitoring of the blend uniformity 33 started from evaluating the blend uniformity of an API with a 34 qualitative technique without chemometric model (Hailey et al., 35 1996; Sekulic et al., 1996). Thereafter, a large number of studies 36 were performed to evaluate the blend uniformity of both APIs and 37 excipients with quantitative techniques based on the chemometric 38 models (Wu and Khan 2009a; Liew et al., 2010). 39 Quantitative techniques are preferable to qualitative in order to 40 judge the blend uniformity with high accuracy. However, 41 developing a robust quantitative calibration model to estimate 42 the blend uniformity with NIR spectra requires enormous labor 43 and time, because the effect of various physical properties such as 44 granule particle size on NIR spectra needs to be considered. In 45 addition, when a calibration model is used for real-time * Corresponding author at: Daiichi Sankyo Co., Ltd., Formulation Technology Research Laboratories, Pharmaceutical Technology Division, 1-12-1, Shinomiya, Hiratsuka, Kanagawa 254-0014, Japan. Tel.: +81 463 31 6954; fax: +81 463 31 6475. E-mail address: nakagawa.hiroshi.w5@daiichisankyo.co.jp (H. Nakagawa). http://dx.doi.org/10.1016/j.ijpharm.2014.05.013 0378-5173/ ã 2014 Published by Elsevier B.V. International Journal of Pharmaceutics xxx (2014) xxx–xxx G Model IJP 14071 1–12 Please cite this article in press as: Nakagawa, H., et al., Verification of model development technique for NIR-based real-time monitoring of ingredient concentration during blending, Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.05.013 Contents lists available at ScienceDirect International Journal of Pharmaceutics journal homepage: www.elsev ier.com/locate /ijpharm