(M1330-09-52) Non-Destructive Quality Control of Medicines Produced through 3D Printing in a Hospital Using Two Handheld Spectrometers

Purpose: Europe’s leading cancer research hospital is conducting a clinical trial with personalised tamoxifen medicines produced via pharmaceutical 3D printing (3DP) for increased treatment adherence [1]. 3DP is a technology for personalisation of medication and small-scale manufacturing for clinical trials [2]. Conventional quality control (QC) measures are unsuitable for the small-batch medicines produced via 3DP [3]. Non-destructive QC methods such as near infrared (NIR) and Raman spectroscopy have been reported as tools capable of assuring the quality of 3D printed medicines. This work sought to investigate the use of two handheld spectroscopic techniques, namely NIR and Raman, for the quantification and QC of tamoxifen in the capsules produced via a pharmaceutical 3D printer in the hospital pharmacy of Gustave Roussy Cancer Hospital for the clinical trial. The two miniaturised spectrometers were investigated for their ability to quantify tamoxifen through the closed capsules or into the open capsule bodies.

Methods: Pharma-ink preparation and capsule filling: Nine pharma-ink formulations consisting of tamoxifen citrate (TC) (20.0 – 40.0% w/w, spaced 2.5% w/w apart) dispersed in polyethylene glycol (PEG) 4000 were prepared for method calibration and two formulations (28.5% and 31.5% w/w TC) were prepared for method validation. PEG 4000 was melted at 78 °C in a beaker and TC dispersed before transfer to 20 mL Luer Lock syringes. Each syringe was heated to 80 °C in the M3DIMAKER2 semisolid extrusion printhead and 100 to 200 mg pharma-ink was deposited in each green/white size 0 gelatine capsule. Four capsules were filled per calibration concentration and three per validation concentration. NIR spectral acquisition: NIR spectra were collected using a MicroNIR 1700ES spectrometer with a tablet probe ( 8 mm diameter collection optic). Open capsules were scanned from above using the NIR printer attachment (ca. 0.5 mm from capsule shell edge, 10 scans per capsule), and closed capsules were scanned through the white capsules bodies (three scans per capsule). Raman spectral acquisition: A Metrohm Instant Raman Analyzer (MIRA) M-1 was used. For the open capsules, spectra were acquired handheld (ca. 1 mm above capsule shell edge, one scan per capsule) into the open capsules through three lens adapters with varying focal lengths (FL); short working distance (SWD, 1.00 mm FL), long working distance (LWD, 7.6 mm FL), and extra-long working distance (XLWD, 18 mm FL) lens adapters. Closed capsules were scanned using a vial lens adapter (three scans per capsule). Data processing and chemometric model development: Tested pre-processing methods included standard normal variate (SNV) and detrend for NIR, baseline estimation for Raman, and Savitzky-Golay (SG) smoothing with 1^st and 2^nd derivatives for both techniques. Partial Least Squares Regression was performed up to 20 latent variables (LV). Characterization and drug loading: The capsule deposits from each concentration of pharma-ink were characterised by DSC and XRD for maintained solid state of TC. HPLC analysis was performed for reference values for chemometric model development.

Results: Quantification of TC from spectra acquisition through closed capsule bodies yielded models of high complexity as observed by the number of LV being 12 and 19 for Raman and NIR, respectively. For NIR, PLSR modelling was performed on spectra pre-processed with SNV and second derivation through SG smoothing (w = 7, p = 2), resulting in high linearity and low calibration error (R2 = 0.996, RMSEC = 0.393% w/w), however a decrease in R2 was observed upon a 10-fold cross validation (CV) (R2 CV = 0.844), further indicating overfitting and non-robustness of the model. For Raman, similar trends were observed (R2 = 0.999, R2 CV = 0.629, RMSEC = 0.0068% w/w). Quantification of TC from spectra acquisition into open capsules was superior when using the MIRA M-1 Raman equipped with the XLWD lens attachment and pre-processing consisting of baseline estimation (smoothness factor (l) = 10⁷, asymmetry factor (p) = 0.05) and SG smoothing (w = 19, p = 2) with 1st derivatisation. A model consisting of five LVs displayed high linearity (R² = 0.998 and R² 10-fold cross validation = 0.963) and low calibration error (RMSEC = 0.293%), indicating a robust model. Moreover, prediction accuracies were only 0.534% w/w TC for the validation capsules whilst 0.707% w/w TC for capsules produced in the hospital pharmacy during the preparation for the clinical trial, indication a very suitable method for TC quantification in the developed medicines (figure 1). XRPD and DSC analyses confirmed the presence of TC in its crystalline form in all produced capsules for the model development and valuation (figure 2 and 3).

Conclusion: Both handheld NIR and Raman spectrometers could yield linear and accurate model for TC quantification in size 0 green/white capsules. The MIRA M-1 equipped with the XLWD lens resulted in the most highly linear and robust model for quantification and proved to be a suitable method for quantification and QC method for the TC capsules produced in the hospital pharmacy for the clinical trial.

References: 1. Denis, L., et al., Developing an innovative 3D printing platform for production of personalised medicines in a hospital for the OPERA clinical trial. International Journal of Pharmaceutics, 2024. 661(124306).
2. Tracy, T., et al., 3D printing: Innovative solutions for patients and pharmaceutical industry. International Journal of Pharmaceutics, 2023. 631: p. 122480.
3. Jørgensen, A.K., et al., Advancing non-destructive analysis of 3D printed medicines. Trends Pharmacol Sci, 2023. 44(6): p. 379-393.

Figure 1. PLSR model (5 LV) for spectra acquisition into open capsules with handheld Raman instrument

Figure 2. XRPD diffractograms of tamoxifen deposits from calibration and validation (*) capsules as well as medicines produced in hospital (**)

Figure 3. DSC thermograms of tamoxifen deposits from calibration and validation (*) capsules