Mixing and Blending Effects: Troubleshooting with process data

Those of us involved in biomanufacturing know the crucial role that buffers (or media) prep/hold tanks or product intermediate holding tanks play in any biopharmaceutical operations. We cannot imagine any biomanufacturing activity without these hold/blending tanks. In large scale operations (any thing above 1000L scale) most of these tanks are stainless steel.  Figure below shows a schematic of such a tank. There can be large variation in the size and design of these tanks within the facility.

Lack of knowledge or consideration of these design variations in operations leads to inconsistent buffer (or product intermediate) preparations leading to incorrect product estimations which in turn leads to incorrect product loading on the next unit operation. Some of the key areas that we should keep in mind are

• estimating appropriate mixing/blending time with factor of safety and incorporating those in relevant SOPs (standard operating procedures) and BPRs (batch production records).
• understanding the relative positioning of regions 1 through 4 (regions are marked by yellow bubble in the tank schematic figure above). Region 2 is important in estimating the concentration (via offline sampling) while region 3 is important in controlling pH (or conductivity) (via online feedback control) respectively. Region 1 and 4 are important in understanding the design differences and relative concentration gradients of solutions within the tank. 

Achieving homogeneity of liquid in these regions (w.r.t to whole mix) is the key to operations with reduced variability and eliminating troubleshooting headaches. Differences in position, type and size of impellers (as shown in figure below) also considerably affect the blending operations which should also be considered while writing SOPs and BPRs.

If you are observing variability (especially in intermediate step protein or buffer concentrations) in your operations it is highly probable that some of it may be coming from inconsistent blending operations. Most of the time blending is overlooked on scale up and we don't invest much time and effort building good data systems around it, hence, the operating staff is not geared with right tool-set for troubleshooting. Usually most of our focus in operations is getting the main unit operations right. We at Simplyfeye highly recommend investing in some extra sensors for online monitoring and detailed recording of manual blending operations for offline historical analysis.

Let me explain this need for data systems with a small example.

Imagine you have a two consecutive chromatography steps ChromA and ChromB within your protein production process. The eluate from ChromA is pooled into a intermediate tank where the pool is blended (and 20% diluted) with a load buffer to prepare for load onto the next step ChromB as shown below

If you had invested in some extra sensors capturing data from tank, the "Tank Weight" trend for a batch on the "Load Prep Tank" would look something like the chart below

The five phases depicted in the chart above (marked by black bubbles) are showing different operations that happened during the blending operations and this data can be sliced and diced accordingly for process troubleshooting:

• Phase 1: Shows tank weight increasing when the eluate is being transferred to the tank. The slope of this curve will give additionally an estimate of the eluate flow rate.
• Phase 2: Shows the hold/mix/idle time before the buffer is added to the tank
• Phase 3: Shows tank weight increasing when the buffer is being transferred to the tank. The slope of this curve will give additionally an estimate of the buffer addition flow rate.
• Phase 4: Shows the hold/mix/idle time after the buffer is added to the tank and a homogeneous mixture is achieved. If the inconsistency is seen in load concentrations this time can be correlated with concentrations 
• Phase 5: Shows decreasing tank weight when the diluted ChromB load is being loaded onto the chromatography column for step ChromB. The slope of this curve will give additionally an estimate of the load flow rate.

As we can make out from this chart above the offline sampling from the tank for estimating protein concentration should be taken

• towards the end of Phase 2 for estimating ChromA eluate pool concentration and
• towards the end of Phase 4 for estimating ChromB load pool concentration

If you have had this information easily available, you could then easily correlate sampling times with the timings of end of Phase 2 and Phase 4 to understand the variability in your ChromA eluate pool or ChromB load pool concentrations respectively. This would help confirm if time of sampling are set properly in the SOPs or BPRs and if not set properly necessary CAPA (corrective and preventive action) can be assigned based on the actual data.

You can also sometime observe a trend where load pool concentration is either getting over estimated or under estimated (as shown in the batch trend chart below), the root cause of which can sometimes be attributed to mixing/blending operations. 

The concentration can be over or under estimated if

• sampling is taken without giving sufficient mixing time
• pH or conductivity control range is too wide
• target concentration calculation is always rounded on the higher (or lower) side of the target
• over/under transfer of buffer due to faulty load scale on either buffer or blending tanks
• line losses in transfer operations
• buffer confounding causing test method error

This explains the need for having an efficient data system that can not only capture but also aggregate the process events, manual operations and machine data efficiently. Otherwise troubleshooting of such problems becomes a nightmare for scientists and engineers working on the floor.

This is just a small example, however, there can be similar other operational effects that are often overlooked giving severe headaches to operations folks in nailing down the variability in their processes. Investing in proper data acquisition and data management tools is worth the money that can save million of dollars in wasted efforts, resources and precious drug product going down the drain. ProcessPad from Simplyfeye is one such platform/solution that can bring some sanity to your process troubleshooting efforts. Write to us if you are solving similar issues in your operations. Our team of experts can help you solve your riddles and also help in setting up efficient data systems.