The Relevance of Continuous Solid Oral Dosage Processing and NIR Spectroscopy In Meeting the Needs of QbD and PAT

In August 2002, the US Food and Drug Administration issued a pronouncement on pharmaceutical manufacturing called Pharmaceutical CGMPs for the 21st Century: A Risk-Based Approach (1). The initiative encouraged the pharmaceutical industry to adopt concepts related to modern quality techniques and state-of-the art processing and paved the way for new thinking about pharmaceutical manufacturing. Current good manufacturing practice (CGMP) guidelines from the late 1980s focused on stasis of conditions based on documented evidence from a small number of validated batches, followed by rigid adherence to standard operating procedures developed for these batches. The modern guidelines are much more dynamic, focusing on a deep understanding of scientific and mechanistic principles associated with the process in order to yield instantaneous and continuous control that can effect change in real time. This represents a paradigm shift from laboratory-centric assurances of quality to one driven by engineering principles.

Two concepts related to risk-based manufacturing are quality by design (QbD) and process analytical technology (PAT). QbD is the more far-reaching term because it involves aspects from development through manufacturing. PAT focuses on manufacturing and relates the mechanistic and scientific interactions inherent to formulation and processing. A necessary connection point for deep process knowledge is the analytical means to establish relationships among the factors of interest. Primary laboratory methods of analysis need well-qualified individuals and substantial time. However, innovations to spectral processing through chemometric models make techniques such as near- infrared (NIR) spectroscopy particularly useful as an alternative. Spectroscopic methods glean information about the interactions between electromagnetic radiation and matter. NIR spectroscopy is grouped within vibrational spectroscopy because it is used for detecting vibrational modes that are associated with functional groups in chemical compounds. Once this technique has been calibrated versus a primary method, the resulting procedure is fast, nondestructive, requires no sample preparation, and is conducive to study in real time. An additional benefit is that the analysis can be automated within standard operating procedures. This makes spectral acquisition and processing covert to the user and merely requires the user to initiate the process with a GO button and await results.

The case for continuous processing in modern pharmaceutical manufacturing

Continuous operations are amenable to process development and real-time adjustment brought about by PAT and lead to continuous improvement. The large established base of batch unit operations associated with solid oral dosage processing may also be studied to elucidate deeper understanding. However, there is a major difference in potential outcomes between batch and continuous processes. Once the batch operation is closed and processing begins, one's ability to correct a quality problem for local (within a mixer or blender for example) or global (to a process before or after the mixer) issues is limited. One may assess content uniformity of the batch in real time, but if uniformity is suspect, the entire batch is at risk. For example, a lubricating step for a direct compression formulation in a bin blender can be studied and modeled to establish appropriate rotation speeds and mixing times to achieve uniformity and minimize particle degradation. However, moving the batch to the tablet press may lead to segregation, which disrupts tablet content uniformity and risks introducing inconsistencies in compression and ejection forces. A batch containing 1000 kg or more of drug product may be at risk. By contrast, a low-throughput continuous-mixing process feeding a subsequent process (continuous rates from 5 kg/h to 100 kg/r or more are feasible) may be monitored constantly at several points. If physical characteristics are trending poorly at the tablet press, adjustments, within the established design space of a well-characterized process, may be made in real time. For example, the rotational speed of the mixer may be increased or decreased to alter the lubricant coating functionality and the at-risk portion of the product run is eliminated or at worse, minimized to the hold-up volume of the mixer.

Continuous processing is characterized by vibrancy, and one can use both feed forward and feed back information that allows interaction among several processes in a global manner. Stryczek et al. (2) provide insight on how the global control potential in continuous operation may be used to influence disparate points in the manufacturing scheme. The study was a retrospective collection of aggregate data related to a solid oral dosage manufacturing process and implicated raw material characteristics, external ambient conditions, and granulating process conditions as influences on dissolution rates. Controlling the granulation process and the other unit operations associated with the process would be a monumental task at the step-by-step batch level involving several manufacturing days with potentially different ambient conditions. Controlling a low-throughput continuous process in a one-day process scheme is much more reasonable because one only needs to consider a single day of ambient conditions.

Continuous processing is adaptable and fits well into the life-cycle approach seen in the 2008 FDA draft guidance on process validation (3). The approach emphasizes the interaction among design, qualification, and verification with the need for continual improvement throughout the product life cycle. The result is a desired state that relies on intimate understanding of critical parameters related to process, product, and quality. Allied with this concept is the potential breadth and depth of the operating space. It is beneficial for a sponsor to include the widest possible range of defined attributes in their application to FDA. Judicious experimental design, factorial sequences, and rapid nondestructive analysis, by NIR spectroscopy for example, make it quite feasible to study many raw material sources as well as wide operating space for equipment variables such as impeller speed, vessel rotation rate, chopper speed, spray rates, and drying temperatures.

Continuous processing is well-suited to process automation schemes. The International Society of Automation (ISA), through their S95 Standards, characterizes levels for integrating control systems with business enterprise systems (4). This is reconstituted in the PAT framework as four increasingly interconnected layers of control and process understanding. At Layer 1, there is some communication of univariate and multivariate process information. Layer 3 is much more elegant and links several analytical devices under common software platforms that are commercially available from several vendors. This seamless interchange can be used to control several aspects within a continuous process. The highest layer encompasses logistics and planning that is well suited to a continuous process where raw-material identification at the receiving area may be matched to final content uniformity assessment, which then triggers automatic reordering of materials when the uniformity is confirmed.

Continuous mixing

A representative unit operation in solid oral dosage processing relates to the mixing of direct compression formulations before delivery at the tablet press. In batch mode, the formulation ingredients would be placed in a vessel and undergo some energy input to achieve uniformity. There are a variety of vessel shapes suitable to the task including round, square, and rectangular bins as well as modifications of the common V-shaped design. Depending on the formulation, additional, more energetic mixing devices within the vessel may be needed. For example, a binary formulation of good-flowing, similarly sized particles may mix quite well while gently tumbling through their angle of repose. In contrast, a formulation with a minor active component and additional functional excipients may need an agitator bar to aid in the march toward uniformity. Various characterizations of classical mixing mechanisms related to diffusive, convective, and shear modes may be studied for deeper knowledge about the processes journey toward uniformity. These issues may be studied with current analytical methods but efficiency suffers with methods that have slow turnaround and destroy the sample.

A secondary mixing step is often necessary for lubricant addition. This formulation component is not amenable to an aggressive mixing environment. A less aggressive mixing environment may also be necessary for lubricating a granulated product that is susceptible to particle-size degradation. In either case, the secondary mixing step may require transfer to another vessel with the potential loss of some drug product, as well as the costs associated with material handling.

Sampling for uniformity, whether of the primary mix or the subsequently lubricated formulation, is often done by thief sampling. Thief sampling is rife with bias points. There is the inevitable drawing down of surface material upon thief entry into the powder bed. The potential also exists for preferentially filling the thief cavity with better-flowing formulation constituents as well as the possibility of smaller particle size components being lost in the annular space of the tubular design. Ultimately, the samples are taken back to the analytical lab for assessment of uniformity. Final release, in turn, depends on the analytical-lab schedule.

The analysis may be done more efficiently by placing an NIR spectrometer on the rotating vessel (see Figure 1) and making the assessment in real time. This method has substantial economic benefits because it is nondestructive and saves quarantine time, but the constraint of this process being in the batch mode limits the potential because of the necessity of adhering to rigid standard opearting procedures. The crucial assessment that one batch is uniform in 15 min versus a 12-min uniformity for a second batch under the same operating procedures is available quickly and nondestructively with NIR, but the ability to quickly use this information to adjust a prior process is impossible within the batch-mode processing regimen.

Figure 1: Thermo Scientific Antaris Target near-infrared analyzer situated on a bin for real-time spectral acquisition. (FIGURE 1 IS COURTESY THERMO FISHER SCIENTIFIC.)

In continuous mode, mixers with highly adaptable throughputs can function quite reliably as pilot-laboratory devices, as well as production devices, using throughput rates between 10 and 100 kg/h. The adaptability is important because, in effect, the same device can be used for development at the pilot-laboratory scale as well as full-scale production. For example, eight hours of processing at 10 kg/h yields the same approximate product mass as a 250-L batch mixer and at 100 kg/h, the yield would be similar to a 2500-L batch vessel.

In the development stage, well-designed factorial experimentation can readily assess the influence of raw-material variations such as particle size and density as they relate to response variables like content uniformity of the effluent process stream. Within the same scope of experimentation, details of variance- reduction ratios and resident time distributions can elucidate mechanistic principles with regard to the feeders and the mixer. The key is that rapid assessment by NIR spectroscopy ensures immediate feedback on content uniformity through nondestructive methods and minimal waste to expensive and limited drug substances. If the continuous mixing device can operate in intermittent mode, as many are capable of doing, there is virtually no low throughput value. In practical terms, continuous throughput rates as low as 5 kg/h are feasible.

A final point is that many continuous mixing devices can be characterized by different mixing regimes within a single device. For example, selection of different flight profiles on the mixing screw can allow the lubricant to be placed in a less energetic section of the device after the mass formulation components have been mixed in a more energetic section. Other mixers have substantial recycling sections allowing the lubricant to be fed into the effluent end of the device through a precise screw feeder, and thus be mixed with other components after the high intensity section (see Figure 2).

Figure 2: Low-throughput continuous mixer processing binary solid oral dosage formulation at 20 kg/h. (FIGURE 2 IS COURTESY HARSCO INDUSTRIALS, PATTERSON KELLEY.)

In production mode, the higher throughput can be engaged under the confirmed process principles determined in development. This abrogates the need to rely on arcane scale-up rules associated with batch mixing. For example, one tries to maintain similarity relationships among vessel sizes based on criteria like geometric, kinematic, and dynamic similarity. The process is amenable to real-time characterization by NIR at several points in the process stream using multichanneled instrumentation (see Figure 3). Fiberoptic probes (see Figure 4) within the feeder hoppers can qualitatively confirm raw-material identification and immediately sense a problem. A secondary confirmation by analytical means within the travel screw can be further related to possible degradation of physical characteristics such as particle size and density. Additional probes along the mixer axis can evaluate content uniformity at several points in the travel through the residence time in the mixer. Ultimately, effluent uniformity for both active pharmaceutical ingredient (API) and lubricant is assured at the mixer exit.

Figure 3: Thermo Fisher Scientific MX FT-NIR Process Analyzer capable of acquiring spectral information on-line and at up to four process points. (FIGURES 3 IS COURTESY THERMO FISHER SCIENTIFIC.)

Any of the attributes that have been sensed can be fed forward and back to effect real time change to other continuous processes. Also, at any point in time, the at-risk portion of the production material is limited to material contained in the hold-up volume of the continuous vessel. Contrast this amount with the at-risk amount in a 2500-L batch mixer.

Figure 4: Diffuse reflectance probe with purge option for analysis of powders. (FIGURES 4 IS COURTESY THERMO FISHER SCIENTIFIC.)

Other continuous-processing operations

Other operations are adaptable to continuous processing and may reap similar benefits if implemented within the QbD and PAT frameworks. The major wet-granulation technologies include fluid-bed granulators, high-shear granulators, and low-shear granulators. Within these product lines, all the technologies offer continuous processing. Opportunities for feed-forward and feed-back optimization with NIR analysis can be related to several steps within the processes. To achieve acceptable wet granulation, some liquid must be added to the dry powder to initiate the growth sequence. The liquid may serve as a binder, activate a dry binder that is part of the granulation formulation, or, in most cases, contain a binder in solution. Fine control of the levels of aqueous or organic liquid addition is possible through NIR analysis because vibrational modes for water and organics are prominent features in NIR spectroscopy, allowing an operator to gauge the amount of solvent in the granulator at any time.

A second control point during the growth phase may be related to gross granule size. NIR is known to correlate to particle size and methods (5) can be constructed to engage a feedback loop for modifying the binder flow rate to maintain a selected particle size range (i.e., 500 to 700 μ). Control strategies related to particle size in a drum granulator have been associated with steady state spray rates of 0.005 kg/s and lower (6). These control strategies are particularly important as they influence yield for start-up and shut-down conditions. Precise definition of binder levels by NIR will minimize waste during these periods.

Continuous drying modes represented by fluidized beds and rotary drying technology may be used as part of the control scheme following wet granulation. Continuous drying can relate to fine control, primarily by using the inherent ability of NIR to sense water. Process-control techniques that allow assessment of the state of drying with regard to the classification of constant, falling, and diffusional periods of drying may finetune the process to minimize unwanted characteristics such as static charge build-up and particle attrition. Spectroscopic differentiation between chemical and free water offers the ability to maintain a multihydrated molecular form of the chemical that is crucial to a subsequent process. Reducing hydration levels often has a detrimental effect on the functionality of a chemical entity.

Continuous extruders are also amenable to control schemes. The extruders have ample process flexibility characterized by the ability to perform multiple operations along the length of the vessel (see Figure 5). Formulation attributes such as improved solubility, reduced dusting, characterization of the API mixing process (dispersive versus distributive), and degrees of crystallinity may be related to constant monitoring of extruder attributes like screw speed, torque, and temperature profile.

Figure 5: Hot-melt extruder with fiber optic probe analyzing active component in extrudate output. (FIGURE 5 IS COURTESY THERMO FISHER SCIENTIFIC.)

Economic benefits

Ultimately, the decision to consider continuous processing is economic, tied intimately to well-designed QbD protocols and PAT implementation as well as meeting the goals defined by lean manufacturing and six sigma quality standards. Waste reduction, reduced analytical expenditures, and faster time to market are prime benefits of a QbD program. Continuous machines can be stopped quite precisely at the end of a run, whereas batch processes are constrained to available batch equipment and thus susceptible to overproduction. Transportation and secondary processing characterized by discharging and charging to several unit operations are mitigated in continuous processing. Analytical results from NIR are available in real time, as opposed to a lengthy wait from analytical laboratories. Finally, a shortened cycle to market and a longer patent life are achievable when the pilot-lab and full-scale production are performed in the same vessel.

A proposed scenario for a hypothetical company that implements lean manufacturing and PAT techniques quantifies the potential (7). The authors show that in the improvements gained, a $450-million company could conservatively expect resulting cost savings of just under $30 million.

Thomas S. Chirkot Ph.D, P.E., is a near-infrared sales specialist at ThermoFisher Scientific, 4410 Lottsford Vista Road, Lanham, MD 20706, tel. 570.909.7657, thomas.chirkot@thermofisher.com

References

1. FDA, "Pharmaceutical CGMPs for the 21st Century: A Risk-Based Approach," (Rockville, MD, Aug. 2002).

2. K. Stryczek et al., "Capitalizing on Aggregate Data for Gaining Process Understanding-Effect of Raw Material, Environmental and Process Conditions on the Dissolution Rate of a Sustained Release Product," J. Pharm. Innov. 2 (1–2), 6–17 (2007).

3. FDA, Draft Guidance for Industry, Process Validation: General Principles and Practices (Rockville, MD, November 2008).

4. ISA, ANSI/ISA S95.00.03, Enterprise-Control System Integration-Part 3: Activity Models for Manufacturing Operation Management (Raleigh, NC, 2005).

5. J. Hirsch and T. Strother, "Lactose Particle Size Analysis Using FT-NIR Spectroscopy," Application Note 51557, Thermo Fisher Scientific, Madison, WI (2007).

6. I. T. Cameron and F. Y. Wang, "Granulation Process Modeling," in the Handbook of Pharmaceutical Granulation Technology, D. M. Parikh, Ed. (Taylor & Francis, Boca Raton, FL, 2nd ed), pp. 555–590 (2005).

7. R. P. Cogdill et al., "The Financial Returns on Investment in Process Analytical Technology and Lean Manufacturing: Benchmarks and Case Study," J. Pharm. Innov. 2 (1–2), 38–50 (2007).