Development of Continuous Additive-Controlled MSMPR Crystallization by DoE-Based Batch Experiments

Additive-controlled crystallization is a promising method to improve crystal morphology and produce solid drug particles with the desired technological and pharmacological properties. However, its adaptation to continuous operation is a hardly researched area. Accordingly, in this work, we aimed to come up with a methodology that provides the systematic and fast development of a continuous three-stage MSMPR cascade crystallizer. For that, a cooling crystallization of famotidine (FMT) from water, in the presence of a formulation additive, poly(vinylpyrrolidone) (PVP-K12), was developed. Process parameters with a significant impact on product quality and quantity were examined in batch mode through a 24–1 fractional factorial design for the implementation of additive-controlled continuous crystallization. These batch experiments represented one residence time of the continuous system. Based on the statistical analysis, the residence time (RT) had the highest effect on yield, while the polymer amount was critical from the product polymorphism, crystal size, and flowability points of view. The values of critical process parameters in continuous operation were fixed according to the batch results. Two continuous cooling crystallization experiments were carried out, one with 1.25 w/wFMT% PVP-K12 and one with no additive. A mixture of FMT polymorphs (Form A and Form B) crystallized without the additive through five residence times (>6.5 h) with 70.8% overall yield. On the other hand, the additive-controlled continuous experiment resulted pure and homogeneous Form A product with excellent flowability. The system could be operated for >6.5 h without clogging with a 71.1% overall yield and a 4-fold improvement in productivity compared to its batch equivalent.

The results of the experiments are summarized in Table S1.After crystallization, the suspension was filtered through a G3 porosity glass filter with a membrane pump and air dried for 3 days.Subsequently, the maximum production was calculated from the weighed mass (see Eq. 2. in the article) and the physical and chemical properties of the product were determined by different analytical methods (see "Experimental materials and methods" chapter in the article).Since 10 and 5 w/t% PVP-K12 caused the forming solid product to stick and coat the glass crystallizer wall, the upper value of p PVP-K12 was fixed at 2.5 w/t% in the fractional factorial design.

Powder flowability classes
In Table S2 the powder flowability classes which were used to characterize each experiment's product are summarized according to the Carr-index.

Product PVP-K12 content examination
The remaining PVP-K12 content of the continuous crystallization product was examined with Raman mapping.A total of 6561 single measurement points per map was an efficient method to detect trace amounts (0.5 w/t%) of solid PVP-K12 dispersed amongst FMT crystals.The spectra were normalized, and linear baseline correction was applied.To enhance the limit of detection in the case of PVP-K12 analysis, the normalized intensity of the reference FMT Form A's and PVP-K12's spectra were adjusted according to their Raman activity.This adjustment was based on the ratio of FMT Form A's peak at 545.6 cm -1 and PVP-K12's peak at 933.3 cm - 1 .The two reference spectra were recorded with the same settings (40 sec spectral acquisition time per spectrum, 2 accumulation number) to maximize the intensity of the peaks and to make them comparable.Examining the additive assisted (1.25 w/t% PVP-K12) continuous crystallization experiment's product, it can be concluded that the sample either contained less than 0.5 w/t% or no remaining PVP-K12 at all.

Batch experiments -Statistical analysis of yield
To check the adequacy of the statistical analysis' results of the fractional factorial design, the residuals were examined.Plotting residuals against case numbers, it can be concluded that randomizing the order of the experiments ensured the independency of errors, as no trend can be seen in the varying of their values (see Figure S1).and the presence of the buffer element (BE) proved to be the statistically significant factors affecting yield.However, plotting the observed yield values against RT and BE shows the difference of functionality, that is if BE is absent from the crystallizer, then yield is a sigmoid function of RT, while when BE is present, then yield changes exponentially (see Figure 4b in the article).Therefore, the analysis of the residuals was done separately for the two scenarios.
Nevertheless, the repetition of the center point experiments resulted small standard deviation in yield, which indicate that the experiments are repeatable, and it is independent of BE.Plotting the predicted yield values (calculated from the fitted equations) against the observed ones, shows that with the BE present at higher yield values, the residuals increase, as the observed yield's standard deviation increase too (see Figure S2a and S2b).This affirms that yield follows an exponential like functionality of RT.However, the number of residuals available might be a limiting factor of the precise evaluation.Nevertheless, based on the normal plot, the errors seem normally distributed as the expected normal values follow a linear trend (see Figure S). the fitted sigmoid model is an adequate approximation (see Figure S3a), the errors are of constant variance (see Figure S3b) and approximately normally distributed (see Figure S3c).
Figure S1.-Residuals vs. case numbers Figure S2.-Predicted vs. observed yield values (a), residuals vs. predicted yield values (b) and normal plot (c) in the presence of BE If the buffer element was not present during crystallization, a sigmoid function describes yield as a function of RT, which is affirmed by the repeated center point experiments.In this case, -Predicted vs. observed yield values (a), residuals vs. predicted yield values (b) and normal plot (c) in the absence of BE

Table S1 .
-Preliminary experiments on the amount of added PVP-K12