HIGH POTENCY APIS
�� () 0 ��
i=
Catalyst [% w / w ] with estimated IPC-limits for impurities
pipeline for faster access and dynamic navigation of process and analytical data . At the core of the initiative was the broadly applicable concept of criticality , which is outlined in further detail below ( Figure 1 ).
An initial version of a criticality-based process data evaluation tool was implemented in 2020 , using Microsoft Excel . While Excel macros support some rather basic data cleaning and evaluation tasks , they clearly do not cut it when it comes to dynamic transformation and the merging of large datasets across multiple dimensions . Even in the context of smaller projects , a significantly more powerful and object-oriented platform is needed to address the data-heavy tasks in a structured , high-performance way . Supported by the very positive feedback from initial customers and users , we decided to migrate the tool to a fully interactive Jupyter Lab data science environment . The new environment , accessed via the web-browser , executes major data transformation tasks on the fly and offers a lot more flexibility when it comes to extending the functionality or adapting the tool to specific needs .
Criticality : A simple but powerful concept
Criticality is the normalisation of a quality attribute against its specification limits : out-of-specification ( OOS ) results have a criticality of > 1 , while results with a criticality of ≤1 are within their specification limits .
In contrast with the raw data , which typically stretch across multiple
�� () 0 �� i=
Catalyst [% w / w ] x - Old setpoint o - New setpoint for process validation
Sweet spot ( all criteria met )
Figure 3 - Reaction DoE evaluation against acceptance criteria for impurities . Note : Estimated , based on IPC purge factors
dimensions and orders of magnitude , dimensionless criticality is normalised to the same scale . This makes it possible to visually summarise and directly compare any numerical quality attribute within the same plot .
When labelled with the corresponding process information , the criticality plot provides an instant insight into the underlying process risks and opportunities . Deviations from typical patterns , or gaps in the data , are easily spotted in the criticality plot , even across a large number of quality attributes .
Using the criticality-based approach , we have successfully established a control strategy and completed a series of very challenging proven acceptable range ( PAR ) studies for a process with more than 100 impurities in a telescoped sequence of organometallic reactions in a flow chemistry set-up ( Figure 2 ).
IPC criticality
The criticality concept is equally applicable to the evaluation of inprocess control ( IPC ) results . It has been successfully applied on IPC purity data in the context of a reaction Design of Experiments ( DoE ). As a first step , an IPC purge model is established by correlating the impurity levels in the isolated product with the corresponding IPC purity data .
Acceptance criteria ( IPC limits ) for impurities in the IPC are estimated by plugging impurity specification limits for the isolated product into the IPC purge model . Evaluation of the reaction DoE against the estimated IPC limits for the relevant impurities indicates the potential for improving the reaction conditions , which cannot be visualised based on the parameters that are typically controlled by the IPC alone ( e . g . conversion , Figure 3 ).
When timelines are tight , IPC criticality helps to prioritise the workup and isolation of experiments based on the IPC purity data . Experiments with a high or untypical IPC criticality usually need to be isolated under standard conditions to assess their potential impact on product quality . Reaction mixtures with a low or typical IPC criticality , on the other hand , are useful as representative input materials to investigate unit operations further downstream .
Since the tool systematically tracks all impurities from the evaluated IPC samples , the probability of overlooking a potential quality risk hidden in the IPC data is significantly reduced ( Figure 4 ). The fact that experiments with criticality of ≤1 meet their specifications implies that any underlying process parameter lies within its PAR . Consequently , PAR ranges can be automatically extracted based on the process parameter ranges covered by the experiments with criticality of ≤1 .
Multiplication of the criticality with the probability of failure for each quality attribute represents a metric for the quality impact or severity inherent to the process parameters tested in an experiment . The tool translates the severity value into a risk priority number ( RPN ), which may be used as a factor in quality risk assessments . Tests in the context of a recently executed PAR study showed that the severity RPN obtained based on criticality was in excellent agreement with the severity levels assigned by an experienced process chemist .
Process data navigation
Criticality provides the high-level compass for in-depth navigation of the underlying process data . The evaluation tool offers a workflow starting with a dashboard that summarises the most relevant high-level project and process information based on criticality
JUL / AUG 2023 SPECCHEMONLINE . COM
17