Aptuit's cGMP API facilities in Kansas City, Missouri and in Oxford, England are equipped to offer chemistry services that span development from mg to 100+Kg and include process research and development and low volume commercial manufacture. Because of Aptuit's scientific strength and expertise in the drug development process, a number of companies have entrusted their drug development programs to Aptuit API teams, from route selection to commercial supply. Trending process data is frequently adopted to identify patterns that help in recognizing potential problems before they occur. This aspect of API manufacturing is an area in which Aptuit excels, enabling the company to offer secure API supply with complete regulatory compliance.

For a commercial process operated at Aptuit's Oxford production facility, a specific stage yield was observed to be in decline but still within the yield specification (Figure 1, see page 2). Since Aptuit had an extensive production history for this product at the site, a wealth of robust data was available to support this apparent downward trend. With full client support and commitment, Aptuit focused on the specific stage yield and highlighted that this yield decline, though small, equated to a reduction of approximately 1.8Kg to 3.0Kg per batch of API supplied to the customer. With a strong emphasis on lean manufacturing and a rigorous commitment to product quality, regulatory compliance and customer satisfaction, Aptuit committed its resources to understand and reverse the declining stage yield.

The Challenge

Aptuit was challenged to understand the cause of the yield reduction and to identify the appropriate measures and controls to return the yield to previously achieved levels.

In chemistry terms, the stage in question was relatively straightforward. It represented a chiral resolution and subsequent recrystallization to provide a product with a defined diastereomeric purity. Since many recorded process parameters were available, the initial challenge focused on data analysis that would allow for a better understanding of how the processing parameters were influencing the stage yield and how these parameters interacted at production scale to achieve the subtle balance of yield versus diastereomeric purity.

Aptuit Solution

With the commercial process based at Aptuit's Oxford production facility, a team of skilled Oxford scientists reviewed the stage processing parameters, providing data on approximately 80 variables from raw material source and purity to process hold times, temperatures, concentrations, etc., through to the stage output yield and purity.

Using a range of graphical tools (including matrix plots, box plots, main effects plot and interaction plots) Aptuit scientists were able to manipulate the data to identify trends within the data set. When viewed graphically, it was clear that the data set comprised two normally distributed subsets and that a statistically significant 'step' change in yield had occurred (Figure 2).

This reduction in yield was confirmed, based on t-test and ANOVA analysis for the most recent 26 batches and was estimated to be in the range 2.1 to 7.2%th with a 95% confidence interval.

Limited evidence could be found for a simple linear relationship between a single processing parameter and yield, though a statistically significant negative correlation between yield and pre-isolation hold time was observed. Contrary to expectation and despite solid stability data, a longer pre-isolation age time at low temperature was found to result in a marginally lower yield. A significant linear correlation was observed between the initial dystomer level (dystomer is a term used to describe the salt of the opposite enantiomer) and the output yield. The stage process involved a classical resolution, the initial output of which contained a well defined level of the dystomer.

A fitted line plot indicated a strong negative correlation between the dystomer levels and the output yields which, scientifically, was expected, since a higher dystomer level can lead to less efficient downstream processing (Figure 3).

An unexpected correlation was also noted between initial dystomer levels and raw material source. Two suppliers had been validated for the provision of the resolving agent and both were able to supply material which comfortably met the associated raw material specification. In fact, analytically, both sources appeared to supply equivalent material. A review of the historical data suggested that material from Supplier A consistently delivered a reaction product with a superior dystomer level when compared to Supplier B (Figure 4).

The data therefore indicated that raw material source was a key factor in the isolation of product with marginally higher dystomer level and that this small increase was sufficient to drive the output yield downwards. Aptuit scientists validated this theory experimentally by undertaking a series of statistically designed experiments (DoE, 2 factor, 3 level full factorial design) and were able to conclude, with a high degree of confidence, that the raw material supplier (and ultimately the raw material quality) was the overriding cause of the decline in yield.

Conclusion

A thorough statistical analysis of the reaction process parameters identified the need for the tighter control of a number of variables including raw material source and reaction hold times. The output from the analysis process allowed the scientists at Aptuit to complete a sequence of designed experiments (DoE) thus allowing those factors important to the observed process variation to be more tightly controlled. Overall, the results were able to drive yield gains, prevent future batch failure and return efficiency gains in terms of reduced cycle times. At Aptuit, it's all about the science..