Synthesis of Vitascor
How to improve yield and Quality
The drug Vitascor is at present produced in pilot scale. Experience shows that the crystallisation step is critical for yield and purity of Vitascor. Yield and purity in this case define the qualities of the end product.
The problem is to determine the relation between the critical process variables and the optimal process conditions in the crystallisation step with respect to yield and purity.
Reliable historical data were not available from the pilot scale production (as very often is the case). A laboratory investigation based on multivariate experimental design for screening purposes was thus necessary. Sirius was chosen for the purpose.
The figure below shows data table, objects are the various weeks and the variables are:
The graph below shows that the Yield varies greatly over the weeks:
With 7 measured process variables, a 27-3 design was chosen. 4 centre points were added to determine noise level, drift and to test significant effects, additivety and linearity in the response models.
The responses (PQ`s) measured for the 20 experiments were %yield (Y1) of theoretical yield of Vitasan and the purity (Y2) which was measured with chromatography. One experiment (12) was not carried out because of an accident.
A bivariate plot of yield vs. purity shows practically the same spread in purity for the centre points (included with R) as the whole experimental series. The conclusion is that purity has no systematic connection with the 7 design variables within the experimental domain.
The spread in yield for the centre points is minimal in comparison to all the experiments. (No drift, no non-linearity's).
A multivariate PLS-response model using Sirius was therefore chosen, between the 7 design variables and the yield.
The result shows in the PLS- Biplot:
The biplot shows that %DMSO and the ratio Hexane/Acetone are negatively correlated to Yield. This indicates that as much as possible of %DMSO shall be removed before the crystallisation and that reduced use of Hexane in proportion to Acetone will also increase the yield.
The other design variables are distributed on an axis orthogonal to the axis through origin and yield and have hence no significant impact on the yield.
This can be tested by plotting the effects (or the regression coefficients) of the design variables in a Normal plot.
The response residuals also plot rather well on a straight line which supports the conclusion that only %DMSO and the ratio Hexane/ Acetone have significant impact on the yield. (The interactions were included in the model, but showed no significance.)
The resulting regression equation is:
%YIELD = 108.1 - 2.15 * %DMSO - 0.41 * HEXANE/ACETONE
It came out that the regression equation above was the best in the investigated area. A plot of measured yield vs. predicted yield showed very good correspondence. A verification of the model by 2 new laboratory investigations and later synthesis in pilot scale showed that a stabile yield of 70% was possible compared with about 25-50% before experimental design was applied.
Why is it not possible to produce 70-100% yield? Because the yield is correlated to residual of %DMSO, the contents of this solution has to be further reduced to increase the yield. Unfortunately, there was limitations in the design of the process itself.
It was not possible to pump slurry with product + solution from the reactor chamber into the crystallisation chamber if the residual %DMSO was below 15%. To further increase the yield, a process had to be constructed where the Hexane/ Acetone is added in the same reactor where the DMSO evaporates.
THE ADVANTAGE OF USING SIRIUS ALREADY AT THE PLANNING STAGE FOR A PROCESS SHOULD BE EVIDENT.