Application of Quality by Design Principles for An Existing Product for the Control of the Dimer Impurity in the Manufacturing Process

Application of Quality by Design principles for an existing product for the Control of the Dimer Impurity in the Manufacturing Process

  Gautam Samanta, Praveen K Sharma, S Srividya, Kamlesh Pai Fondekar, K R Janardan Sarma, Saurabh Gupta, GRB Ravish, and Y Raghuram   Dr. Reddy's Laboratories Integrated Product Development Organization Bachupally, Hyderabad India

The philosophy of quality by design is all about the patients' need. The designing, developing, and manufacturing of the product must meet those needs. Most of the innovator and few generic companies have adopted the concept of QbD in their product development. Dr. Reddy's Laboratories has started implementing the QbD principles for all new APIs development. The QbD implementation for new products is well documented but not for existing products. The key drivers for the implementation of QbD for the existing products are to reduce variability in the product quality, improve yield, reduce cycle time, solve manufacturing issues, reduce cost of quality, and introduction of real time release testing for manufacturing process.

The objective of this work is to provide a perspective of QbD application for existing API manufacturing where a stepwise reverse QbD approach has been applied. The unique features of this work comprise methodology and implementation strategy in defining design space and thereafter executing the control strategy for consistent production of the API.

Irbesartan is an angiotensin II receptor antagonist used mainly for the treatment of hypertension. Dr. Reddy's Laboratories developed the generic version and filed DMF in March 2009. In stage 1, IRB-1 is formed by reaction between 2 n butyl-4-spiro cyclopentane-2-imidazoline-5-one and 4-bromo methyl-2'-cyanobiphenyl at -5 to -10 °C in the alkaline medium (Figure 1). The reaction time is 2-3 h. In stage 2, IRB-1 is converted to tetrazole by the reaction with NaN₃ at elevated temperature and the final product Irbesartan (IRB-2) is obtained after crystallization.

Figure 1: Stage 1 for the synthesis of Irbesartan and the structure of the Dimer impurity (CQA)

 In stage 1, there is a significant variation (upto 5%) in the formation of a dimer impurity, which is difficult to remove in the subsequent stages. If the dimer content is more than 1%, significant yield loss occurs for its removal. So the cost of quality is high. The dimer impurity is the critical quality attributes (CQA) in this stage. The QbD principles have been applied to address the issue.

The following stepwise QbD approaches has been adopted:

1.      Historical data and analysis of data to identify the problem

There is a significant variation in the formation of the dimer in the process. So some variables are playing critical role. To understand the variability in the dimer quantity experiments were conducted at pre-determined time under the same experimental conditions.Figure 2 shows the experimental results, which indicate that the formation of the dimer depends on the reaction time. Even within the reaction time 2-3 h, the dimer quantity is varying significantly and the variation is more after 3 h which indicates that reaction time is very critical. The best fit with the data is the third-order polynomial equation (R²= 0.994).

Figure 2: Best fitting of the experimental data

2.      Quality risk assessment for material attributes and process parameters

To identify the critical variables the cause-effect diagram (Figure 3) has been used. All material attributes and process parameters have been listed in this diagram. Based on the chemistry knowledge as well as practical experience it has been concluded that material attributes are not responsible for the dimer. So only the process parameters have been considered for risk analysis.

 

Figure 3: Cause-effect or fishbone diagram for MA and PP to understand the effect on CQAs

In the case of PPs, the first phase is to identify those potential CPPs that may affect the CQAs of the API (i.e., the severity). The second phase of the assessment is to estimate the probability of a parameter affecting the API (i.e., the risk). The parameters with a high probability of influencing the API with enough severity to produce unacceptable API are critical parameters. The overall risk is the combination of severity and probability and risk assessment is the impact of the process on the API. Table 1 shows the risk assessment of the process parameters.

Table 1: Risk assessment for the process parameters

Based on the preliminary risk analysis of the process parameters, RPM, NaOH quantity, and reaction time have been identified as critical process parameters. Though the reaction temp is critical, it has not been considered for DoE study because the reaction is not completing above -5 °C and the reaction mass is freezing below -18 °C. Due to a narrow window of operation it was decided to fix it at -7.5 ± 2.5 °C. To establish the relation between the critical process parameters and CQAs the design of experiments (DoE) has been applied.

3.      DoE, understanding of process, and design space

Initially, two-level factorial design was selected and twelve experiments (2³ + 4 center points) were conducted in the lab considering three variables and their respective ranges as mentioned in Table 2. The normal plot for the dimer (Figure 4) reveals that all the three variables are making significant positive impact on the dimer formation (NaOH 36%, RPM 21%, and Reaction time 13%). It is interesting to note that the three-variable interaction (ABC) has negative impact (15%) on the dimer formation. Due to significant curvature a central composite design was selected and remaining experiments were conducted.

Table 2: Critical process parameters and their ranges

Figure 4: Half-normal plot for two-level factorial design

The data from CCD are fitted to a cubic model (R² = 0.9999, Adj R² = 0.9997, Pred R² = 0.9496, and PRESS = 2.76). Analysis of variance testing (Table 3) confirms the significance of three-factor interaction (ABC), as well as of other terms A²B, AB², A²C, A², B², C² and two-factor interactions AB, AC, and BC. The three-dimensional surface plot, illustrated in Figure 5 is generated according to the cubical model and demonstrates the dependence of dimer on the amount of NaOH and RPM at optimal level of reaction time (2.5 h). To see the influence of three-factor interaction (ABC), the model term (ABC) has been excluded and the analysis of variance indicates that in the absence of that term almost all the terms become insignificant (Table 4) and the lack of fit becomes significant. This indicates that the three-factor interaction term is playing very critical role in the formation of dimer.

Figure 6 is an overlay plot of yield, dimer, and other QAs, setting the foundation for continued improvement of the process. The yellow region is design space where the desired QA levels will be obtained. Laboratory validation experiments were performed under the experimental conditions (NaOH 1.3 eq, Reaction time 200 ± 10 min, RPM 300 ± 20) chosen from the design space and the dimer (0.29 ± 0.06) and other quality attributes were within the specifications.

Table 3: ANOVA table and associated statistics for response surface cubic model for dimer

Figure 5: Influence of NaOH and RPM on the formation of dimer impurity at reaction time 2.5 h

Table 4: ANOVA table and associated statistics for response surface cubic model for dimer excluding ABC interaction term

Figure 6: Overlay plot of Yield, starting materials, purity, dimer impurity, and single maximum unknown impurity for stage 1, which provides a process operating window

4.      Control strategy

Does the output from the laboratory-scale multivariate design get reflected during scale-up? It is important to do the appropriate scale-up for the scale dependent variables before manufacturing. Based on the process understanding and criticality a control strategy has been developed to consistently ensure product quality. To control the formation of the dimer in the plant the scale dependent variable, RPM was scaled-up based on tip-speed and control on other variables was achieved through automation. In-process analysis was not available to monitor the formation of the dimer at that time. The consecutive six batches data in the plant indicates that the dimer is under control within the specification (<1%).

The QbD principles provide a structured approach for gaining process knowledge and developing robust manufacturing process. The application of the QbD principles for the existing products is very limited. This has been applied successfully to reduce the variability of the dimer impurity in the process and reduces the cost of quality. On the basis of the process understanding, it has been demonstrated that the control of the dimer impurity can be achieved by controlling the critical process variables.