Incorporating Process Analytical Technology Data into Process Control

A Q&A with Ivo Backx, Siemens Industry Automation Division
Mar 20, 2013

Process analytical technology (PAT) is being successfully used to improve understanding and optimize pharmaceutical unit operations, but greater value can be obtained by integrating PAT with overall process control of a continuous manufacturing system. Pharmaceutical Technology spoke with Ivo Backx, manager of business and project development for the pharmaceutical industry at Siemens Industry Automation Division, to gain insight on the issues involved.

PharmTech: What is involved in using the data from PAT measurement to control a pharmaceutical process?

Backx: Regulatory authorities are pushing the pharmaceutical industry to better understand the manufacturing process and to measure the attributes that are critical to the final quality of the manufactured product, with focus on the therapeutic effect of the medicine. Critical to quality attributes (CQAs) must be measured and monitored against deviation limits. Parameters that are critical to the process (CPPs) must be controlled in order to keep the CQAs within the design space and continuously produce the right product with the right quality.

In the SIEMENS PAT quality data management system (SIPAT), the first step is to collect all relevant data from different sources, such as multivariate data from spectroscopy instruments as well as univariate data from classic instrumentation. The spectrum from an analyzer can be read out over a direct interface. Univariate data is typically already available in the process control system and can be collected from there. Data from off-line laboratory tests may be of interest as well and can be obtained over an interface to a laboratory information management system (LIMS). Principally, data from the different sources must be collected and integrated. In this process, time alignment is very important as the data are collected at different sample rates in the different systems. Context data (e.g., equipment identification, product-plug identification, batch or lot identification) are retrieved from the process control system and added to the data, thus allowing for product tracking and process data mining.

Statistical data regression models are applied to the collected data using chemometrics software runtime engines. The result is a value of the CQA (e.g., moisture content or particle size). The derived CQA can then be aggregated to calculate standard deviations and averaged values and can be monitored over time in a trend display. The aggregated values of the different CQA data over different unit operations of the total manufacturing process are the basis for advanced process control for closed-loop, model-predictive control, which steers the CPPs in the process control system. The different calculated CQAs are stored electronically and are the basis for real-time release evaluation.

PharmTech: How are these systems being used today in finished solid dosage processes (e.g., tableting)?

Backx: Today, PAT instruments are often used on single unit-operations to measure a CQA. Often, they are applied for closed-loop control but isolated to a single process step. For instance, moisture content is measured to determine the end point of a drying cycle. Hardness and thickness are measured and controlled in the tablet press. The concentration of the active ingredients may be verified basing on NIR analysis of tablet samples after the press.

In order to get in-specification product the first time and all the time, however, the control strategy must be to take all the CQAs into considerations and deploy a feed-forward/feed-backward process control strategy over the complete manufacturing process.

PharmTech: Are there any other issues that limit the use of PAT in process control?

Backx: Today, PAT instruments still have dedicated, proprietary interfaces. Now a new interface standard, the OPC UA ADI (Object Linking and Embedding for Process Control Unified Architecture for Analytical Device Interfaces), exists, but few instrument suppliers support it.  Another problem is that spectroscopy instruments are quite expensive. They are typically applied in laboratory environments and require more robustness to support a commercial manufacturing. This situation, however, is changing fast as new instruments come on the market. Another issue is that multidisciplinary teams are required to develop and maintain the PAT solutions, which is a challenge for most organizations.

Process equipment suppliers often have a single PAT solution integrated in their equipment controls, with local data management on the local equipment human-machine interface. Validation of isolated, single operation solutions is a pain; tighter integration is a must. Platforms that integrate all CQA data from different sources (i.e., online, inline, at-line and off-line) into one central quality data management tool and deploy the overall control strategy from a central system should be used.

Ivo Backx is manager of business and project development for the pharmaceutical industry at Siemens Industry Automation Division,

lorem ipsum