Purpose: To develop an in-line NIR method to determine the residual acetone, moisture and potency of a spray dried intermediate during the secondary drying process in a real time fashion.
Methods: To construct the NIR models to predict the residual acetone, moisture and potency of the spray dried solid dispersion intermediate (SDDi), a simplex-centroid design was utilized to establish the boundraries of SDDi compositions. The SDDi consists of active, polymer and additive at the range of 18-24, 8-14 and 68-74 % (w/w), respectively. The solid components were initially dissolved into an acetone-water mixture at 3 to 1 ratio at a total solid loading of 10% (w/w). These solutions were spray dried using a pilot scale spray dryer (PSD1, GEA) at batch sizes of ~400g total solid/formulation under the same spray drying conditions. Subsequently, several scale-up batches (~6000g of total solid/batch) of SDDi solutions with the same solvent and solid loading were prepared. These solutions were spray dried using a production scale spray dryer (PSD2, GEA) using different outlet temperatures and spray rates to generate SDDi materials with different properties.
The secondary drying process was conducted using either a 3L (~400g/batch) or a 100L (~6000g/batch) Ekato agitated dryer under reduced pressure with an agitator tip speed of 0.8 m/sec. The product temperature was controlled via the jacket temperature of the dryer vessels and ranged from 37° to 43° C. To enhance the flowability of the wet SDDi out of the Ekato, 1.5% of SiO2 per batch of wet SDI was co-charged with the SDDi into the Ekato dryers for secondary drying.
Diffuse reflectance spectra were obtained with an Antaris EX FT-NIR spectrophotometer (Thermo Fischer Scientific, Waltham, MA) equipped with a 1” OD diffuse reflectance probe in the dryer. The software package RESULT by Thermo Fisher Scientific was used to collect the spectra. For the offline reference methods, HPLC, Gas Chromatography (GC) and Karl Fischer (KF) were used to determine potency, residual acetone and moisture, respectively.
Quantitative analysis by means of Partial Least Squares (PLS) regression was performed with TQ Analyst software by Thermo Fisher Scientific for data analysis with spectra pretreatment.
Results: In this investigation, all ~400 g batches were used for the NIR models building. Two of 6000 g batches were used for the model update (training set) and the rest of the 6000 g batches were used as the testing set. Quantitative NIR models for acetone, water and active assay were built directly with the spectra collected from the 3 and 100 L Ekato dryers. The correlation plot of residual acetone predicted by NIR and measured by off-line GC with all batches is displayed in Figure 1. The NIR data correlated well with the offline GC analysis acetone with a standard error of calibration (SEC) of 0.064 %wt.
The correlation plot of residual water predicted by NIR and measured by off-line KF with all batches is displayed in Figure 2. The NIR data correlated well with the offline KF analysis of water with a SEC of 0.061 %wt.
The correlation plot of potency predicted by NIR and measured by off-line HPLC with all batches is displayed in Figure 3. The NIR data correlated well with the offline HPLC analysis of potency with a SEC of 1.78mgAPI/gSDI.
The NIR models showed that under the Ekato drying conditions of the study, the SDDi drying process reached its target end-points (Acetone <3000 ppm and water <1.5%) in less than 5 hours regardless of the “wetness” of the starting materials. The drying curves predicted by NIR were confirmed via offline HPLC, GC and KF tests.
Conclusion: Online NIR methods have been built to quantify acetone, water and API during the secondary drying process in a real-time manner. A SEC of 0.06 wt% and a Standard Error of Prediction (SEP) of 0.11 wt% were obtained when the NIR method was used to predict the acetone level at the end of secondary drying of different batches. Similar results were obtained for water with NIR measurement (SEC 0.06 wt% and SEP 0.18 wt%). A SEC 1.78 mgAPI/gSDI and a SEP of 3.0 mgAPI/gSDI was obtained when the NIR method was used to predict the potency of active. Overall, these results indicate that NIR methods can be built as alternative in-process tests of residual acetone and water.