Process Optimization: From Petrochemicals to Biopharmaceuticals

By Thomas Daszkowski, Sanjay Joshiand Peter Ryan, Ph.D. and P.E., Bayer Technology Services

Introduction

For many years, Bayer Technology Services (BTS) has been developing and applying optimization methods and tools to processes operated by its internal customers. Initially, its optimization efforts focused on improving Key Performance Indicators (KPIs)—throughput, specific energy consumption, color, yield, off-gas quality, and product purity among them—for Bayer’s large, continuously-operated isocyanate and polymer plants in Baytown, Texas, and in Leverkusen, and Dormagen, Germany.

BTS based its methodology on Six Sigma’s stepwise DMAIC approach, to Define, Measure, Analyze, Improve and Control the process. A standardized process-monitoring tool was developed and used in each of the DMAIC steps of the process-optimization project.

Themonitoring system collected process data from various process databases, including distributed control systems (DCS), laboratory information management systems (LIMS), analyzer databases and Aspen simulationdatabases. The system then assessed process KPIs and their

associated process variables. Once the KPIs were established, more-advanced statistical analysis tools were used to develop each KPIs set of associated process variables more fully.

At this point, process optimization proceeded to the “Improving” phase, and the monitoring tool was then used to plan and observe process trials. Finally, as the project reached its final “controlling” phase, the monitoring tool served to sustain the results of the optimization project.


As the tools and methods were developed for optimizing processes in these facilities, it became clear that the same tools could also be used to optimize Bayer’s batch and semi-batch pharmaceutical and biotech processes. Despite their obvious differences, semi-batch biotech processes and large continuous polymer processes shared some basic features:

Table 1: Commonalties Between the Chemical and Biotech Sectors

As shown in the table, each requirement on the continuous (polymer) side has a counterpart on the semi-batch biotech side. Reactor kinetics, for example, can be associated with cell metabolism, reaction with fermentation, and distillation with chromatography. The commonalties found in this table served as the basis for using the methods and tools developed in the chemical and polymer sectors to optimize the batch and semi-batch processes found in the pharmaceutical and biotech sectors.

Case Study 1: Computer Simulation in the Chemical and Biotech Sectors

Computer simulation is a key part of BTS’ approach to process optimization. The mathematical representation used for the simulation may be “fundamental,” or based on a set of equations, or it can be “data-based,” based on the plant’s performance history. Depending on the types of processes involved, simulations can be steady-state or dynamic.

A well-developed process simulation is essential in order to understand the relationships between the process parameters associated with a KPI. This need for understanding is just as apparent in a large continuous polymer process as it is in a small batch or semi-batch biotech process.

Comparisons may be drawn between the requirements for modeling a chemical reactor and those for modeling the human body. In order to optimize the reaction conditions in a chemical reactor, a dynamic model of the reactor must be developed to predict how the reaction will progress with varying temperature, pressure, reactant concentrations, and catalyst concentrations.

Developing this dynamic model requires proposing a reaction mechanism based on the physical properties of the reactants and products and bench-scale batch experiments. Dynamic simulation can then predict the concentration profiles of reactants and products over time as the optimization variables are changed. If the model produces concentration profiles that agree with the bench-scale experimental results, the proposed reaction mechanism is accepted. If, on the other hand, the model cannot pass its concentration profiles through the set of experimental data points, the reaction mechanism is modified until the model’s performance improves.

Similarly, in order to optimize the delivery of a new drug into the human body, a dynamic model of the human body can be developed to predict how the drug will be distributed and metabolized, and finally excreted from the human body. BTS has developed one such “pharmacokinetic” (PK) simulation called PK-Sim, which is based on mathematical representations ofthe drug’s distribution through plug-flow transport, diffusion through a membrane, and metabolism using Michaelis-Menten kinetics.

The model’s parameters are developed using in-vitro and small-animal experimental data. The dynamic simulation can then predict the concentration profiles of the drug over time while the characteristics of the drug (for example, its solubility, lipophilicity, or molecular weight) are changed.

If the model produces concentration profiles that agree withexperimental results, the proposed distribution mechanism is accepted. If, on the other hand, the model cannot pass its concentration profiles through the set of experimental data points, more detail must be included in the model; for example, the model may have to consideractive transporters in the lipid membrane.

Reactor simulations are developed to improve knowledge of the reacting system. With the simulation results, a pilot reactor can be designed, and the economics of the process based on the optimized reaction mechanism can be examined before the pilot facility is even built. In the same manner, the PK-Sim® simulation is developed to gain knowledge for the drug absorption and distribution characteristics in the human body. With the PK-Sim® results, comparisons between the new drug and existing drugs can be made in terms of distribution and metabolism in the human body, and a clinical trial strategy can be developed. The economics of the new drug in terms of concentration profiles and accumulation of the drug in the targeted areas of the body can be examined before the initial clinical trials are started.

Case Study 2: Total Performance Optimization in the Chemical and Biotech Sectors

There is a real need for process optimization in pharmaceutical and biopharmaceutical manufacturing. As shown in Table 2 below, the cost of quality, as measured by the number of defects in a process, is significantly higher in the pharmaceutical sector than is other process industries such as the chemical and semiconductor sectors.



Table 2: Performance of a Process in terms of Cost of Quality[1]

In addition, the return on investment (ROI) for optimizing a biopharmaceutical process can be much higher than for other processes, as summarized below in Table 3.

Table 3: Typical Optimization Project ROI

As shown in Table 3, a shorter pay-out period is expected with a chemical sector project because fewer regulatory issues are encountered, but a significantly higher ROI is typically associated with the capacity or yield increase of a pharmaceutical or biotech project because of the margins of the product.


Proceeding with the capacity increase example, the process variables associated with the KPI shown in Figure 1 (capacity increase) must be identified, and the interactions between these process variables must be established. The identification and analysis of the process variables is accomplished using statistical analysis tools such as principle component analysis and partial least squares. Figure 2 shows example results of such a statistical analysis. The results show that of the process variables considered as effecting the KPI, only a small number of these variables have a significant impact on the KPI, with some having a positive impact and others having a negative impact.

Using these results, an optimization proposal is developed which can include modification to operating procedures, improved control, and tighter scheduling of the process step. The emphasis at this stage of the optimization project is to avoid or minimize the capital investment costs of the project while working within the validated process parameters.

Since yield and capacity increase are high-priority KPIs of a biotech optimization project, determining the baseline capacity of the unit is a critical first step of a Six-Sigma based optimization project. Once a baseline production capacity is established, based on historical unit performance, simulation results, review of equipment specifications, validation documentation and standard operating procedures, the variances of the KPIs found in the process can be analyzed. Figure 1 demonstrates the results of a variance analysis for a process step in a batch biotech facility.

Having developed and implemented the plan presented in the optimization proposal, a series of process trials is carried out to verify that the targets outlined in the proposal have been achieved. The optimization project can then focus on sustaining the improvements demonstrated during the process trials. Sustainability tools are now applied to the process which continuously monitor the control system. If a control loop is not in the intended mode, or if the loop becomes saturated or oscillatory, the loop is flagged and an alarm is displayed on the control monitoring screen. Using such sustainability tools, the optimized process is held at its best demonstrated performance level.

The example optimization project described here must accomplish its targets and goals within the constraints of a validated process. The tools and methodologies described in this example are similar to the tools and methodology described in Guidance for Industry, PAT - A Framework for Innovative Pharmaceutical Manufacturing and Quality Assurance, US FDA, August, 2003.Recall that in this document, PAT refers to Process Analytical Technology. The guidance document describes the tools available to improve and optimize a process using a PAT approach; the PAT tools include[2]:

  • Multivariate data acquisition and analysis tools,
  • Modern process analyzers or process analytical chemistry tools,
  • Process and endpoint monitoring and control tools, and
  • Continuous improvement and knowledge tools.

The example presented above makes use of each PAT tool to realize the targets set forth in the “Define” stage of the optimization process.


Figure 2: Results of Statistical Analysis of Historical Data

Conclusions

It can be clearly stated that commonalties exist between the chemical and polymer sectors and the pharmaceutical and biotech sectors. The methods and tools used to optimize processes in the chemical and polymer sectors have been developed and used with success for many years. Based on the success of these tools and methods, and on the commonalties between these two sectors, BTS has focused on, and realized similar results in the biotech sector. The tools and methods described above and applied to large, continuous processes were readily adapted to the smaller, batch and semi-batch processes found in the pharmaceutical and biotech sectors. The most challenging aspect of optimizing biotech processes has been realizing the targets of the optimization project within a set of validated process parameters.

Bayer Confidentialpage1 of 5

[1] FDA Science Board Meeting, November 16, 2001, D. Dean, F. Bruttin, Price Waterhouse, modified by Ajaz S. Hussain, Ph.d., Deputy Director, Office of Pharmaceutical Science, CEDR, FDA

[2] Guidance for Industry, PAT – A Framework for Innovative Pharmaceutical Manufacturing and Quality Assurance, Draft Guidance, August, 2003, Page 10.