A Risk-Based Approach to Product and Process Quality in Spray Drying

May 1, 2008
Pharmaceutical Technology, Pharmaceutical Technology-05-01-2008, Volume 2008 Supplement, Issue 2

Process designs and control strategies can be improved by adopting a risk-based approach to product quality. The author describes how this approach can be applied to spray-drying operations.

Taking a risk-based approach to product quality is an excellent way to uncover potential weaknesses in process designs or control strategies. A risk assessment is different from an impact assessment because the former considers the severity and the likelihood of the impact. A properly performed risk assessment is a valuable tool in designing experiments, defining design space, and planning design-of-control strategy. The ultimate goal of risk assessment is not only to achieve the desired quality, but also to produce predictable and consistent quality.

PHOTO COURTESY OF NIRO.

Target product quality and product-quality consistency

The target product quality in this instance may be the desired product quality for full-scale production or a quality that is sufficient for the current scale and stage of development. Product-quality variability is a common problem that can be difficult to solve. Product-quality risk assessment is an effective tool to reduce product-quality variability.

Quality impact assessment

The first stage of performing a risk assessment is identifying potential hazards. A hazard is something that influences the process directly or indirectly (e.g., materials of construction or instrumentation). A typical impact assessment will identify hazards, the type of hazards, and the documentation required such as material or calibration certificates.

Impact assessments are typically designed to identify potential hazards, but not the severity or frequency of exposure to the identified hazards. Since neither the severity nor the frequency of the hazards is identified, an impact assessment is inadequate for designing future experiments or control strategies.

Quality risk assessment

Risk is a combination of the severity of exposure to a hazard and the frequency of being exposed. Being struck by lighting is a good example of a low-risk hazard. Despite the severity of the hazard, the likelihood of it happening is small.

A product-quality risk assessment requires both product and process understanding to be performed effectively. Consequently, it is best performed after process development has been completed. Ideally, the target product quality is known at that point, including the extent to which the different product-quality attributes can vary without unacceptable quality loss. The limits of acceptable variability are rarely known for all product attributes, however. A quality risk assessment can often identify the aspects of the process that require further investigation. Quality risk assessment must be included in the experimental design or receive special attention in the control strategy.

It is important that personnel have adequate general understanding of the process type (in this case, spray drying), especially if the process will be scaled up at a later stage.

Applying control strategy

An ideal control strategy would allow large input variation but maintain low output variability. Unfortunately, these criteria are rarely met in the real world. A good control strategy would allow some input variation, maintain low output variability, and provide early warning of deviations. The success of such a control strategy is ultimately limited to known input variation. Process analytical technology (PAT) is a valuable tool in identifying input variation that previously was overlooked or deemed insignificant (because of multivariate effects, for example).

The control strategy can be simplified if the product-feed variations are small, as they frequently are. In this case, a good control strategy is to fix the process parameters that have the greatest product-quality impact by controlled variations in the process parameters with the least product-quality impact.

Spray drying

In a spray dryer, a liquid feed (e.g., solution, suspension, or emulsion) is atomized into a spray of fine droplets and suspended in gas while drying. Spray drying is a fast method of drying because of the product's large surface area and high heat-transfer coefficients. The product's large surface area also enables drying at low to moderate temperatures. The rapid drying and consequent fast stabilization at moderate temperatures make spray drying feasible for heat-sensitive materials.

Spray dryers must always be cleaned and sometimes sterilized. These and other subprocesses are often as critical as the main drying process, but are not part of the assessment as described below.

Identifying the important process parameters

In a spray dryer, droplets or particles are dried while they are suspended in the drying gas. During the drying process, heat is transferred from the drying gas surrounding the droplets or particles, and solvent evaporates from the droplet or particle surface into the surrounding drying gas (see Figure 1). The process also involves various complex diffusion processes, most of which are specific to the product–solvent combination used. This complexity makes liquid-feed composition an important part of early process development. Because the liquid-feed composition rarely can be used for process-control purposes, however, the parameters available for process control are drying temperature, drying-gas humidity, and droplet size.

Figure 1

The conditions under which the droplets are transformed into particles influence the final solvent content and the particle morphology.

Drying temperature

In a spray dryer, it is important to distinguish between dryer-inlet temperature, product temperature, drying temperature, and dryer-outlet temperature. In a well-designed cocurrent spray dryer, the intense mixing of spray and drying gas results in fast cooling of the drying gas by the evaporation of solvent. The end result is that the temperature in the drying chamber is practically equal to the spray-dryer outlet temperature. This parity can be measured and calculated with computational fluid-dynamic models (see Figure 2).

Figure 2

Because of the evaporation of solvent from the product, the droplet or product temperature remains lower than the spray-dryer outlet temperature during the entire drying process. The droplet or product temperature also remains lower than the spray-dryer outlet temperature in the high-temperature region at the drying-gas inlet, where the rate of evaporation is at its highest and the droplet temperature is at its lowest (i.e., approaching the wet-bulb temperature). The product temperature at the discharge is typically between 5 and 20 °C colder than the drying-gas temperature at the outlet.

The importance of the drying-gas temperature at the spray-dryer outlet is clearly seen in experiments where the correlation between product characteristics and outlet temperature is stronger than most other process variables. The outlet's drying-gas temperature is usually fixed at a product-specific set point by a feedback control loop. The feedback loop normally adjusts the inlet's drying-gas temperature or the liquid-feed rate to maintain the outlet's drying-gas temperature at the set point (see Figure 3).

Figure 3

Drying-gas humidity

The solvent-vapor content in the drying gas is the sum of the solvent-vapor content in the inlet drying gas and the evaporated solvent. In most applications, the evaporated solvent is the major contributor. Because the evaporation rate is proportional to the difference between the inlet and outlet drying-gas temperature, an increase in the inlet drying-gas temperature (at constant outlet drying-gas temperature) results in an increase in drying-gas solvent-vapor content.

The control of the inlet's drying-gas temperature depends on the control strategy applied for the outlet's drying-gas temperature. When the outlet drying-gas feedback loop adjusts the inlet's drying-gas temperature, the fixed set point for the liquid-feed rate indirectly sets the inlet's drying-gas temperature level (see figure 3). Alternatively, a feedback loop maintains the inlet's drying-gas temperature at the set point by adjusting the main process gas heater.

The solvent-vapor content in the inlet drying gas is controlled effectively and accurately by adjusting the condenser's outlet gas temperature (i.e., the dew point of the gas). In applications that use ambient air as the drying gas, the condenser is replaced by a dehumidifier or sometimes completely omitted.

Droplet size

Droplet size is controlled through the atomization process as long as suspended particles in the liquid feed are small compared with the droplets created. Atomization is a complex but reproducible process in which droplet size and droplet-size distribution depend mainly on the rheology of the liquid feed, the energy applied, and the liquid-feed rate. The atomization energy can be applied and controlled in various ways such as adjusting the speed of a spinning wheel in a rotary atomizer, the flow of a gas in a two-fluid nozzle, and the liquid-feed pressure in a pressure nozzle (see Figure 3). For a given pressure nozzle and liquid feed, the liquid-feed rate and the liquid-feed pressure are mutually dependent. Either of them can be used in a feedback loop to the feed pump to maintain constant atomization conditions and thereby constant droplet size and droplet-size distribution—as long as the liquid-feed rheology remains constant.

Process-gas flow rate

Spray dryers are generally designed to work correctly within a limited range of drying-gas flow rates: typically ± 10–20%. The process parameters that have been selected as critical above, however (i.e., inlet's drying-gas temperature, outlet's drying-gas temperature, and feed rate), are sufficient to calculate the heat and balance across the system. The drying-gas flow rate is thus a dependent process parameter, not a variable process parameter.

Though the drying-gas flow rate is a dependent process parameter, it is impractical to control it according to real-time mass-balance and heat calculations. A simple feedback loop between a gas-flow measurement and the main process gas fan works just as effectively. Gas-flow measurement in this case does not require an accurate absolute measurement: a reproducible relative measurement also suffices.

Product-quality risk assessment

Assessments of risk are relative. A risk's magnitude is gauged in comparison with other risks, and "high risk" is relative.

Process measurements and control loops. In most spray-drying applications, a strong, reproducible correlation exists between product quality attributes that are influenced by the process parameters and the basic process measurements (e.g., temperature, pressure, and flow). Basic process measurements are reliable. Regular calibration and preventive maintenance reduce the risk of deviations even further. The undetected failure of a process measurement or control loop is not likely. Because the process parameters must satisfy the heat and mass balance, an undetected error in one instrument would cause other process measurements to deviate from the normal values.

Process measurements and control loops are effective and reliable process controls and constitute a low risk.

A few spray-drying applications do not exhibit an adequately reproducible correlation between the product-quality attributes that are influenced by process parameters and the basic process measurements. In those cases, the process measurements and control loops are not effective for controlling the process and constitute a high risk.

Variation in feed characteristics. Dried products' characteristics change with liquid-feed variations. The process impact of long-term variations in feed characteristics can be difficult to evaluate because the varations are complex and frequently multivariate. One problem is that a limited number of raw-material batches are used in product development, and some sources of variability are easily missed or deemed insignificant.

The combination of control loops that rely on constant response to constant conditions and the difficulty in detecting and analyzing variations in the liquid feed make liquid-feed variations a moderate risk.

In cases with simple liquid feeds (e.g., a solution of one simple chemical compound), the liquid-feed variations become a low risk.

Mechanical errors. Mechanical failures and assembly errors (e.g., missing or damaged gaskets) may cause contamination, deposits, or malfunctions that are not readily detectable. Operator training, standard operating procedures, and preventive maintenance reduce the risk of error, but a high risk of mechanical errors remains.

Applying control strategy

Process measurements and control loops. The process measurements and control loops are effective and reliable. A real-time heat and mass-balance calculation based on the process measurements is used to detect instruments that need calibration or certain liquid-feed abnormalities.

Spray-drying applications with an inadequate correlation between product-quality attributes and the basic process measurements require careful reevaluation of the process design space. The lack of correlation is frequently explained by the choice of a marginal or unsustainable operating point.

For example, an operation close to the limits for drying often results in particle agglomeration in the drying chamber and a high risk of deposits or irregular powder discharge (see Figure 4). Furthermore, single particles rarely have the same performance as agglomerates of the same size. The result is a poor correlation between atomization conditions and particle size because the drying conditions are marginalized typically by lowering the outlet drying-gas temperature.

Figure 4

Fine powders and fragile particles may change size dramatically when they are collected in a cyclone. Cyclones have a limited efficiency in collecting fine powders. As a result, a reduced particle size may appear as a reduction in cyclone yield and not as the expected reduction in particle size. On the other hand, fragile particles may break when collected in a cyclone. In extreme cases, attempts to increase the particle size through changes in atomization conditions make the particles even more fragile. Again, the result is a poor correlation between atomization conditions and particle size. The lack of correlation appears to stem from the drying-gas flow rate. The real cause, however, is a poor choice of the cyclone for the application.

Product may change after discharge if the conditions in the product container are not compatible with the product. When the product changes, it reflects the process conditions in the container, not in the spray dryer. Some product characteristics to monitor are particle size (some products are likely to form lumps or agglomerate), residual moisture (reabsorption of solvent vapor because of the increase in relative humidity as the surrounding gas cools down), amorphousness (crystal growth in products kept at temperatures above glass-transition temperature), and activity or impurities (product is kept at an excessively high temperature for too long).

Variation in feed characteristics. The process impact of long-term variations in feed characteristics can be evaluated using PAT. In line or on line liquid-feed and final-product monitoring are ideal for establishing a correlation between feed properties and final-product properties. Unfortunately, it is time consuming to set up the monitoring system and analyze the collected data.

A fully developed liquid-feed and product-monitoring system can be used as an advanced feed-forward or feedback system to adjust operating parameters. Such a system can reduce the consistency requirements for the liquid feed and, at the same time, improve the consistency of the final product.

A system with only in line or on line final-product monitoring still requires a liquid feed with a high level of consistency. Even though, in most cases, variability is best controlled at its source, such a system still provides benefits in processes with a tight design space or in a development environment where real-time data allow the process to be adjusted quickly. The danger in such a system is when the cause of the disturbance is not known.

The choice of measurement type and location is naturally application-dependent. An in line or on line measuring device is generally able to provide substantially more detailed information than traditional sampling. For example, an on line particle-size measurement shows the effect of the automated hammers on the drying chamber (see Figure 5).

Figure 5

Mechanical errors. Mechanical failures frequently are not part of a process-control strategy but are one of the most common reasons for process deviations. A pressure nozzle, for example, must be clean and without leaks to work as intended. Pressure nozzles, however, often start to leak because they have been damaged, worn out, or assembled with insufficient care. The leak causes the nozzle to foul, disturb the spray, change the particle size, create deposits, and ultimately cause a premature shutdown. Monitoring and recording nozzles with a camera provide early warnings and facilitate fault identification.

Summary

A risk-based approach to process design provides valuable insight to the areas where process design and control strategy are most likely to fail. Interestingly enough, it is frequently not the high-risk areas that personnel devote most of their attention to before making a risk assessment. In many cases, the most effective process monitoring is achieved though a blend of technologies.

Subprocesses such as cleaning can and should be evaluated in the same way as the main process and can be improved by means similar to those described above.

Henrik Schwartzbach is a senior process technologist at Niro A/S GEA Pharma Systems, Gladsaxevej 305, 2860 Soeborg, Denmark, tel. +45 39 54 54 54, henrik.schwartzbach@geapharmasystems.com