CompareNetworks
Applying QbD Principles to
Lyophilized Drug Product Development
Dr. Sajal Patel: --Sajal Patel, and I am in the Formulation Sciences, Biopharmaceutical Development there since I started. And what I am going to present today is how we can apply the QbD principles to lyophilized drug product development.
Can you hear me well? Okay, thanks.
And a quick outline of the presentation. I will start with an introduction. I'll briefly talk about what do we mean by a QbD approach for lyophilized drug product development and then talk a little bit of formulation and process development challenges and consideration, and lastly talk about the application of QbD, and specifically the PAT process analytical tool and design space, and conclude with a summary.
So, lyophilization is a problem coupled in formulation and process. One really cannot separate the formulation from the process, and they really need to be studied together.
Understanding the formulation is critical for successfully designing and developing a robust lyophilization process with acceptable product quality attributes, whereas understanding the freeze-drying process itself along with the container closure system, the freeze dryer unit, is critical for scaling up and optimizing the process.
So, what do we mean by a QbD approach for lyophilized drug product development? So, it starts--basically QbD aims at building quality within the process rather than monitoring offline at the end of the process. And for freeze-drying process, understanding the impact of formulation and process performance on drug product quality, it's really critical to really apply the principles of QbD.
And the QbD begins all the way from pre-formulation characterization and specifically the thermal characterization of the formulation in regards to the TG prime or TC. TG prime is the glass transition temperature of maximally freeze-concentrated solution, and TC is the collapse temperature. And really, these parameters define the process parameters that one can use during the freeze-drying process development.
The next step is the cycle development technologies such as SMART freeze-dry, TDLAS, which is tunable diode laser absorption spectroscopy, and several PAT tools that are now available to monitor and control the freeze-drying process.
These technologies really help to reduce the cost of the batch as well as the development time to develop the freeze-drying process. And this results in increased process understanding for optimization and eventual scale-up from lab to production scale.
The next step is process robustness or process characterization. And several mathematical tools are now available to monitor or to actually model the different steps of the freeze-drying process. And there are several publications in the literature applying the QbD approach for freeze-drying process development.
And really, this mathematical modeling needs to be used to gain the process understanding and also develop the ability to assess the impact of excursions, like temperature and pressure excursions, that happen on a production scale freeze-dryer.
And the last one is post-lyo product characterization. So, extensive characterization of the lyophilized product itself is critical to understand the impact of formulation and process performance on product quality attributes and product stability.
So, the next few slides are example of--emphasizing pre-lyo formulation characterization. So, why is this important and what is the significance of thermal characterization of the formulation?
So, typically the primary drying step of the freeze-drying process is performed at--by keeping the product temperature below the maximum allowable product temperature, TP max. And this TP max is usually TG--is less than TG prime or TC for an amorphous matrix. And TC is about a degree or two higher than the TG prime.
And in general, drying above TG prime results in cake collapse, which further impacts product stability. However, for high concentration protein formulation, visible collapse may not be observed on the timescale of the experiment. And if this is really true, then this opens up the opportunity to dry above TG prime, resulting in shorter lyo cycle time and high throughput.
In the instance where one would like to stay conservative and conduct the primary drying below TG prime, having the data to--wherein the drying is carried out above TG prime without impact on product stability really helps to support temperature and pressure excursions that happen on the production scale unit.
So, what we have in this slide is the dataset for two formulation matrices. The one on the left is for sucrose and the one on the right is for trehalose. What we have on the Y-axis is temperature, and on the X-axis we have the protein concentration in micro mil.
Let's just focus on the dataset for sucrose. The square symbols represent the collapse temperature measured by freeze-dry microscopy. The triangles represent TG prime measured by DSC. And what we see is that, up to about 45, 50 mg per mil, TG--TC is within a degree or two of TG prime. However, above 50 mg per mil, we start seeing the significant difference between TG prime and TC.
And independent of the formulation matrix, whether it's sucrose or trehalose, the qualitative trend remains the same. And this is important because every one degree C increase in product temperature during primary drying results in about 13 percent reduction in primary drying time.
So, clearly you can see the importance of thermal characterization for the freeze-dried formulation that, instead of TG prime, if one were to use TC as the maximum allowable product temperature, one would easily result in a 60 to 70 percent reduction in primary drying time and the total lyo cycle time.
So, freeze-drying is a problem coupled in heat and mass transfer. And the key objectives during freeze-drying process development is to obtain a stable and elegant product with minimal inter and intra batch variability, and process should be easily scalable and transferrable from one lyophilizer to another. And at least in industry, one additional objective is to minimize the processing cost.
And in order to achieve these objectives, understanding the critical process parameters, CPPs, is really important here. Now, freeze-drying is based on fundamental principles of heat and mass transfer. And by no means I am going to walk over the theory of heat and mass transfer.
But, the point here is that there are several mathematical models available to model the different steps of the freeze-drying process. But, these models are as good as your input parameter. Garbage in will be garbage out. So really, even to use these models, you really need to understand and characterize your formulation, primary container closure system and--as well as the freeze dryer itself in regards to heat and mass transfer to get any meaningful data out of these mathematical models.
What we have in this slide is some commonly encountered freeze-drying process development and scale-up issues and some considerations. By no means it's an--a comprehensive slide, but a very few selected ones which are key especially during cycle development on a lab scale freeze dryer.
So, the first one is a freezing step wherein we have heterogeneity in the ice nucleation temperature. Now, ice nucleation is random and stochastic in nature. So, we--because of the differences in ice nucleation temperature, we have wide variability within a batch as well as batch to batch variability, depending on the environment wherein the freeze-drying is carried out.
And there are technologies now available to control this important parameter during the freezing step. And one such technology is available from Praxair ControlLyo which enables controlled nucleation at the desired temperature.
One could argue that, "Why not just include an annealing step and take care of this heterogeneity during the freezing?" However, if you introduce an annealing step, then defining the annealing temperature and time is really critical. And if you have crystallizing excipients, then of course you do need an annealing step to ensure that the excipients do crystallize during the freezing step.
The next one is the edge vial effect, or atypical radiation effect. So, any vial that is not surrounded by six other vials is defined as an edge vial. And this vial receives higher heat transfer, primarily via radiation, because the door and the walls of the freeze dryer are at relatively higher temperature compared to the shelf itself.
While most of the vials in the freeze dryer are center vials, it's important that this edge vial consideration is taken early during development to ensure that not only the center vials but also the edge vials meet the desired product quality attributes.
The next one is determination of end point of primary drying. Why is this important? Because if the drying time is too short, you will prematurely progress into secondary drying, resulting in either collapse during via glass transition or melt back due to presence of ice. If it's too long, then now you have a conservative lyophilization cycle unnecessarily occupying the dryer time.
The last one is the effect of dryer load. And in a nutshell, as the batch size increases, product temperature decreases, drying time increases. And the next few slides is actually a case study emphasizing the importance of the batch size in a freeze-drying process development.
So, this case study is for a novel molecule, S, and it's a lyo scale-up case study for a Phase III clinic supply. And we have an efficient lyophilization cycle, about 43 hours, for this molecule with appropriate safety margin built in.
We made 17 clinical lots with this lyophilization cycle on one of our production scale freeze dryers, what I am calling here as Dryer A, at less than 40 percent of its capacity. However, when Phase III scale-up was required, a two to three X increase in the batch size was required, which would push either running Dryer A at 100 percent of its capacity, or we move on to Dryer B at 50 percent of its capacity.
So, some potential risks were identified before we actually executed the clinical batches. A, we were talking about running Dryer A at its full capacity and, as I said, that we have an efficient lyophilization cycle of about 43 hours, which results in rapid sublimation rate during the early part of the primary drying step. And B, now we are talking about running the cycle on Dryer B, on which we have no prior experience running the cycle parameters.
So, the hypothesis that--was that either the primary drying time may not be sufficient for a full dryer capacity or even 50 percent of dryer capacity and we may have collapse or melt back, or we may have loss of process control either via choked flow or condenser overload.
So, choked flow is a phenomena wherein the speed of gas in the duct connecting the chamber and the condenser reaches the speed the sound, wherein the flow in the duct gets choked. As a result, the water vapor cannot condense onto the condenser coils, but rather it gets backed up in the product chamber, resulting in an increase in product chamber.
So, the challenges identified were how do we assess the process performance and product quality impact on either running Dryer A at 100 percent of its capacity or Dryer B at 50 percent of its capacity? A and B are, again, production scale freeze dryers.
So, we started our exercise by primary drying time analysis, and that's what we have in this chart. We have drying time on the Y-axis and batch size and percent of full load on the X-axis.
The current primary drying time is 14 hours, and that's the red solid line here. Those different datasets that you see are for Dryer X, Y, Z. X and Y are lab scale freeze dryers, whereas Z is more sort of a pilot scale freeze dryer.
And what we see is that, as the batch capacity increases, drying time increases. However, this is for lab and pilot scale freeze dryer. But, the question we are asking is for a production scale freeze dryer, Dryer A and B.
So, based on the data from the lab scale freeze dryers and based on some limited data from the production scale freeze dryer and using some mathematical modeling, we estimated the drying time on Dryer A and B at different batch size.
And what we have here is that, at 100 percent of full capacity on Dryer A, which is the square symbols, we will be running close to the current primary drying set time of 14 hours without much safety margin, whereas on Dryer B definitely we won't have any safety margin built in. We will be right at 14 hour mark. However, at 50 percent of the batch capacity on Dryer B, we would have some safety margin still built in in the process.
So, overall, the analysis was made that there is not much potential risk for not completing primary drying and progress prematurely into secondary drying, at least with the current drug product presentation as well as the cycle parameters.
The next concern we had was a choked flow or the--a condenser overload. So, in order to evaluate this, what we did is we loaded one of our pilot scale freeze dryers, Dryer Y, at its full capacity. And we estimated that the average sublimation rate under this condition will be about 273 grams per hour with a maximum of around 340 grams per hour.
And we used the dryer characterization data, which is acquired by just running simple sublimation tests on the freeze dryer where you load trays of water and freeze them, and then ran the shelf temperature from minus 40 to whatever is your desired shelf temperature, let's say up to plus 20 degrees C.
The chamber pressure is set to zero millitorr. And under the sublimation rate that is achieved at the shelf temperature, the chamber pressure will equilibrate to certain steady state value. And that is defined as the minimum achievable chamber pressure in the chamber, product chamber.
And that's what we have on the Y-axis, PC min. And on the X-axis, we have the sublimation rate. So, what we did is, using this set of data, we said, okay, 341 grams per hour. What is the minimum pressure that this unit can achieve? And we estimated it to be about 200 millitorr.
However, the current set point for the cycle is 150 millitorr. So, even without doing the experiments, this data suggested that we will not be able to achieve a chamber pressure of 150 millitorr in the cycle.
And actually, we decided to go ahead and do the experiment, and that's the data on the right-hand side. And, you know, pressure on the Y-axis, time on the X-axis. The red border line is the chamber pressure set point of 150 millitorr. The blue curve is the absolute pressure in the chamber measured using capacitance manometer. And the orange curve is a Pirani gauge.
And what we see is that indeed we were never able to achieve the set point of 150 millitorr and, rather, the pressure went all the way up to about 200 millitorr, as we had estimated from the dryer characterization data.
However, this is for dryer--the data that we presented is for Dryer Y. But, again, the question we are asking is for Dryer A and B, which are the production scale freeze dryers. And that's the comparison we have in this graph.
We have PC min on the Y-axis and sublimation rate on the X-axis. Diamonds and squares are Dryer X and Y. They are real similar in design and geometry, so you can see that the PC min lines are really well with--for both of them. However, Dryer Z is more sort of a pilot scale freeze dryer. And very luckily, we had two data points from one of our production scale freeze dryer, the orange symbols.
And the orange circles line up really nicely with Dryer Z. And when we did an analysis of the design and geometry between these two freeze dryers, indeed they are very similar in regards to their product chamber, condenser design, and well as the duct connecting the chamber and the condenser.
So, the assumption was that Dryer A is very--is going to be very similar to Dryer Z. And what we see is that, at 100 percent of the capacity on Dryer A, sublimation will be around 2,000 grams per hour, or two kg per hour, and we will be very close to that choked flow limit and we will have an issue with chamber pressure control.
And for Dryer B at 50 percent of the capacity, again, we didn't have any data points to compare for Dryer B, and it was very different in design and geometry. But, based on some mathematical extrapolation, we estimated that, even on Dryer B at 50 percent, we will have some risk of running into process control issues.
So, independent of the dryer, it was at least concluded that running 100 percent of the batch size even on Dryer A and Dryer B is really risky in regards to process control. So, a risk-based decision was made to split the batches and run at 50 percent of the capacity independent of the dryer, either Dryer A or Dryer B.
So, this exercise clearly emphasizes the importance of dryer characterization and how such data can be used to perform not only process--during process development but also scale up between the--as well as--scale up from one lyophilizer to another lyophilizer.
And especially this is important for organizations that do optimize their processes. And this is important not just to ensure the product quality but, from a company perspective, it's important to make the maximal use of your lyo capacity.
So, just changing gears now, I'll talk a little bit about PAT tools, process analytical tools, to monitor and control the freeze-drying process. So, what this table shows is the different steps of the freeze-drying process, freezing, primary drying, secondary drying; and what are the important physical properties during the steps and what are the PAT tools that are currently available on the market; and the potential impact on the product quality attributes such as residual water, recon time, cake appearance, and physical stability, primarily aggregation.
And the point of this slide is that really monitoring and controlling the critical process parameters, the CPPs, is an integral element of quality by design. And the next few slides at least demonstrates a case study, primarily how to monitor the endpoint of primary drying, a step of the freeze-drying process.
So, what we have in this graph is the Pirani pressure profile in pink, pressure reading on the left-hand Y-axis, residual water reading on right-hand Y-axis in percent, and time on the X-axis.
The pink curve is the Pirani pressure profile. The aqua curve is the chamber pressure set point, which is 60 millitorr in this case. And underneath the aqua curve is the blue curve, which is the actual pressure in the chamber measured using capacitance manometer.
The red diamonds are residual water determined by gravimetry, weight before minus weight after. And the circles are residual water determined by Karl Fischer. And the case study is for 5 percent sucrose, which is representative for an amorphous matrix. Similar data were also obtained for 5 percent mannitol, which is representing a crystalline matrix.
And what we see is that, as drying time progresses, residual water decreases. At the onset of drop in Pirani pressure, residual water is about less than 20 percent. And at offset, where Pirani pressure is almost equal to the CM, capacitance manometer, residual water is less than 5 percent.
So, clearly a sharp drop in Pirani pressure indicates that primary drying is about to complete. What it does is that the gas composition is changing from mostly water weight, but up here, to mostly nitrogen down here towards the end of primary drying.
So, in this study, what we did is we compared several different techniques to determine the endpoint of primary drying, and we have two tables here. The table on the left is for sucrose. The table on the right is for mannitol.
And the techniques that we evaluated is the Pirani gauge, the data that we just saw in the earlier slide; Lyotrack, which is a gas plasma spectroscopy; condenser pressure wherein pressure is measured using a CM in the condenser; dew point, essentially measuring the gas composition in the product chamber; a product thermocouple, it's a traditionally used technique to determine the endpoint of primary drying; TDLAS, tunable diode laser absorption spectroscopy, which essentially is a technique that is mounted in the duct connecting the chamber and the condenser, and one is measuring the concentration of water vapor in the duct. And the last one is a pressurized test, or MTM, essentially measuring the vapor pressure of ice in the product chamber.
What we--the numbers in the table are the percent residual water, and the color indicates the cake appearance. Red indicates that we had a melt back due to presence of ice. Yellow indicates we had cake collapse via glass transition. And green indicates we had good cake structure retention.
Let's just focus on the table on the left, which is for sucrose. And what we see is that the condenser pressure, product thermocouple, TDLAS, pressure rise test, they all result in high residual water and melt back at the onset.
And techniques such as condenser pressure and product thermocouple also result in high moisture and melt back at offset. Only techniques like Pirani, Lyotrack, and dew point resulted in good cake structure retention with the relatively low residual water.
However, it's important to mention here that in these experiments the samples were pulled out of the product chamber and then quickly equilibrated at room temperature, and hence we saw the collapse.
However, when one were to progress from primary to second drying, one would use a relatively slow ramp rate, typically .3 degrees C per minute and low--lower, to ensure that we don't see the collapse that we observed in these experiments.
And also, for an amorphous matrix such as sucrose, the amount of unfrozen water at the end of primary--at the end of freezing step is about 20 percent. So, many of these samples that are color coded yellow and with about 20 percent residual water can still be used to determine the endpoint of primary drying.
For mannitol, which is a crystallizing excipient, as expected pretty much all the technique resulted in good cake structure retention with relatively low residual water.
And the conclusion of this study was that Pirani by far is the best method to determine the endpoint of primary drying. It doesn't mean that other techniques don't work. Rather, other techniques like TDLAS and MTM not only determines the endpoint of primary drying, but they also provide several other useful information that can be used for process optimization as well as scale-up.
The next application of PAT for freeze-drying is the ice nucleation temperature, which essentially governs the pore structure that is formed in the freeze-dried cake. And as I mentioned earlier, ice nucleation is random and stochastic in nature. And it is a critical process parameter because every one degree C increase in ice nucleation temperature results in about 3 percent reduction in primary drying time.
And what we have in this graph is specific surface area on the Y-axis and several different formulation conditions on the X-axis. Formulation one through five had controlled nucleation via ice pore technique, whereas formulation six is uncontrolled ice nucleation.
And the point to make in this slide is the error bars. If you look at the error bars for controlled nucleation, they are very tight compared to the uncontrolled nucleation where we have almost more than an order of magnitude difference in error bar, suggesting we have more heterogeneity in the cake morphology for uncontrolled nucleation samples versus the controlled nucleation. And techniques such as SEM really helps to do a qualitative assessment of the cake morphology in regards to the pore structure.
And really, by controlling the ice nucleation temperature, we can at least minimize and to a certain extent eliminate the inter batch and intra batch viability arising due to differences in the ice nucleation temperature, and also it helps to eliminate the scale-up issues arising due to differences in ice nucleation temperature between a lab scale freeze dryer versus production scale freeze dryer in class 100.
So, what is the end goal of all the systematic approach of formulation characterization, process development, optimization, and scale-up? Where are we going with all of this?
So, the end goal is really to establish the operational design and control space. And the example that we have here is the primary drying step of the freeze-drying process. So, one of the input parameters for the primary drying step is the chamber pressure, and an important output is the mass flux. That's what we have on the Y-axis.
And this white open space is what we are defining as an unexplored space, unexplored because for a new formulation, for a new presentation on a new lyophilizer, we don't know how it's going to behave.
However, when you start putting constraint on your input parameter, for example for pressure, there is--on production scale freeze dryer it's difficult to control chamber pressure below 60 millitorr. Then why design a process with chamber pressure set point of 60 millitorr and below?
So, that's where we put an--a limit on the lower set point for chamber pressure. For chamber pressure higher than 300 millitorr, there is not much added advantage in regards to freeze-drying process. So, we put the upper limit at 300 millitorr.
And within these two red lines is where we defined the knowledge space. This is the knowledge space because a lot of knowledge exists within the company, within the literature, within the public domain how to design and develop a freeze-drying process with a chamber pressure set point of 60 to 300 millitorr, and then you start imposing the freezer dryer capabilities and limitations.
And the example here that we have is a choked flow limit. It could be a condenser overload limit as well. And that's the dotted line here, the choked flow limit. And the regime above this dotted line is defined as choked flow regime, meaning that you don't want to be designing your process wherein you are somewhere in this regime. Otherwise, the process will be out of control.
So really, you want to design your process in this space, and that's what we define as the operational space for that given lyophilizer. And knowing this operational space early on is really important, to make sure that when you are scaling up the freeze-drying process from lab scale to production scale, you don't run into the issue that we showed in the batch size case study that we had, where on a lab scale at 100 percent you will be fine. But, when you scale it up to a production scale, you run into an issue of process control, and you are limited now on the batch size that you can run on your lyophilizer.
One of the other important parameters for the freeze-drying process is the product temperature. You want to make sure your product temperature is below the maximum allowable product temperature, which could be the TG prime or TC for most of the freeze-dried formulation.
In this example, we have the limit of minus 25 degrees C. And if you have a limit of minus 25, you want to set a lower limit. In this example, we set it to minus 30 degrees C because, if you run it more colder, you just have a conservative process. So, you should be designing your process between these two parallel lines.
But, one doesn't control product temperature directly in a freeze-drying process. It's actually the shelf temperature and chamber pressure that indirectly controls the product temperature.
We have the chamber pressure on the X-axis. So, let's put the shelf temperature isotherms, and that's where we have--the triangles are shelf temperature set point of minus five, whereas the diamonds are shelf temperature set point of minus 15 degrees C.
So, in order to achieve a product temperature of minus 25 degrees C, you can either run it--run the process here, which is about minus five degrees C and 140 millitorr or 130 millitorr, or you can run the process here, which is minus 15 degrees C and about 240 millitorr. So, somewhere in here is your design space wherein you should be designing your process.
And of course the control space should be within this design space. And the set point should be defined such that you have enough safety margin built in on either side for shelf temperature and chamber pressure.
So, that's the end goal of systematically applying the QbD principles to define a design space. An example here is the primary drying step of the freeze-drying process.
Also, additional consideration should be given to the physical characterization of the product during the freeze-drying process as it changes its physical state from frozen, partially frozen to partially dried, dried, and freeze-dried.
And there are several physical properties that are important during each of these steps, and there are techniques that can be used to characterize a product during these different steps. Again, the primary product quality attribute that it impacts is the residual water, recon time, cake appearance, and physical stability.
So, just to summarize, formulation and process development challenges really needs to be taken into consideration during early development to ensure product quality. Formulation characterization tools that I didn't touch base a lot in the previous slide really helps to provide a macroscopic and microscopic comparative analysis to determine if any physical changes has occurred in the product. And these tools are really versatile when changes are made either to the site, to the scale, to the freeze-drying process parameter, or to the drug product presentation.
And PAT tools really confirm that the process is running as expected and it's--and assuring that the product quality is built within the process rather than monitoring offline at the end of the process. And PAT and design space are integral element of quality by design.
With that, I would acknowledge all my colleagues at MedImmune who helped put this presentation together. And thank you for your attention. I am ready to take the questions. Thank you.