Aqueous Cleaning and Solvent Substitution in Chemical Synthesis API Manufacturing

Most active pharmaceutical ingredients (APIs) are manufactured by chemical synthesis in which fine chemicals and intermediates undergo significant chemical change through a series of multistep processes. These synthesis processes typically include the use of organic solvents and, therefore, traditionally require organic solvents for process cleaning. A growing trend exists in the industry to move away from solvent-based cleaning to aqueous cleaning whenever possible, driven by safety, regulatory and economic factors.

The recent release of the International Conference on Harmonization (ICH) Q7A GMP Guidance for APIs has generated significant worldwide interest and discussion concerning the validation of API manufacturing processes. However, limited published information is available regarding the design of an efficient and validatable cleaning process for API manufacturing.1

Traditional solvent-based cleaning

Synthetic API manufacturing involves many pieces of equipment for reaction and separation processes. Typically, these include reactors, condensers, crystallizers, centrifuges, distillation or extraction columns, filters, dryers and associated pipes. The nature, level and tenacity of the residues encountered in the process equipment may vary widely even within the same production train. Cleaning reactors, where aggressive and prolonged processing conditions are encountered, may pose a different challenge from the cleaning of separation equipment such as centrifuges or dryers, in which slurries or solids may be caked onto the surface. A variety of materials such as glass, polytetrafluoroethylene (PTFE), hastelloy, stainless steel and polymers are used in the construction of process equipment. A good cleaning process will take all these variables into account to ensure that all surfaces that come into contact with the product are safely and effectively cleaned.

The traditional approach to cleaning has been to use an organic solvent, usually the same process solvent used in the synthesis of the API. This approach focusses mainly on the solubility of the active ingredient and frequently ignores the varying effects of the solvent on different surface types, the chemistries involved and the processing conditions previously mentioned. The most widely used cleaning solvent in the industry is methanol; other commonly used solvents include acetone, dimethyl formamide and ethyl acetate.

The cleaning process often starts upstream with the introduction of the cleaning solvent into the reactor. Reactor capacity can range from 50–500 gal for a pilot plant to more than 3000 gal for large-scale manufacturing processes. Solvent-based cleaning typically involves agitating a solvent in the reactor vessel, circulating it through pipes, and refluxing the heated solvent through an overhead riser and condenser system. Refluxing condenses the solvent vapours on overhead vapour lines and condensers, thus allowing for wetting of the surfaces and possible residue removal by dissolution in the solvent. Although this type of cleaning is similar to the processing step in a reactor and, therefore, has advantages, it is often very slow and gradual because of the limited level of "action" or mechanical force on the various overhead equipment

surfaces. For this reason, it is not uncommon for as many as 5–10 solvent boil-outs to be required to clean some tenacious residues.

Advantages of organic solvent cleaning

• The API is usually soluble in the organic solvent. Solubility data are often easily accessible.

• The solvent may be readily available and routinely used in the manufacturing process.

• Solvent residue analysis is simple, and may be unnecessary if the cleaning solvent is the same as the process solvent in the next batch.

• Coverage of overhead surfaces is possible by solvent vapour refluxing. This allows difficult-to-reach areas to be cleaned without using spray devices or modifying equipment.

Disadvantages of organic solvent cleaning

• Residues other than the active ingredient (such as degradants) may be present on the surface and are not soluble in the cleaning solvent.

• The solubility of a solvent does not necessarily indicate the rate of solubility. Vapour refluxing is a time-consuming process. The action or force acting on the surface during this process is very limited. The residue also has the potential to redeposit in overhead pipes and on surfaces as the solvent evaporates.

• Solvents are flammable and may require special consideration for equipment design, storage, handling and transportation.

• Environmental issues associated with solvent-based cleaning processes can be significant. Solvents such as methanol are hazardous air pollutants that require compliance with specific regulations.2 Applicable regulations in the US include 40 CFR Parts 9 and 63 National Emission Standard for Hazardous Air Pollutants (NESHAP), and the pharmaceutical Maximum Achievable Control Technology (MACT) standards. Water effluent limit guidelines are covered under 40 CFR Parts 136 and 439.

• Discharging large amounts of cleaning solvents may be an issue with corporate and community images. In addition, spent organic solvents are usually either recovered or incinerated, thus adding to the overall cost of manufacturing. For this reason, several pharmaceutical corporations have set goals for solvent reduction.

• Not all surfaces in an API process train can be easily cleaned without manual intervention. Organic solvents can be toxic by inhalation, so special safety procedures and protection may be required.

Aqueous-based cleaning

Aqueous-based cleaning processes are the norm in finished pharmaceutical and biotech industries. Two broad categories of aqueous cleaning agents are used in these industries: single-component commodity cleaning agents such as sodium hydroxide or phosphoric acid; and formulated cleaning agents, which are multicomponent cleaners that take advantage of several cleaning mechanisms.3

Although commodity cleaning agents are widely used in the biopharmaceutical industry because of their simplicity and ready availability, they have limited soil-suspending and cleaning ability for the tenacious residues typically encountered in the API industry. Formulated detergents take advantage of the synergy of various components in the formulation. Like commodity alkalis, but unlike organic solvents, a formulated cleaner can include alkalis to enhance solubility and help in the hydrolysis of the residue. Formulated detergents can also contain surfactants that provide better wetting, surface action and emulsification, chelating agents that break down complex metals such as calcium and iron, and dispersants that prevent particles from reaggregation.

Table I: Evaluation of detergent chemistry.

However, vapour refluxing is not usually an option with aqueous cleaning agents as it is with organic solvents because of the non-volatile nature of the active ingredients in most aqueous cleaning agents. If aqueous cleaning agents were refluxed, then only the water would vaporize, and the cleaning effectiveness in the reactor dome, vapour lines and condensers would be minimal. Aqueous cleaning, therefore, requires direct liquid spray or recirculating flow coverage across all surfaces.

Advantages of aqueous-formulated detergent cleaning

• Aqueous formulations are not flammable, which reduces storage, transportation and handling costs.

• Aqueous systems can be more effective at cleaning because they can incorporate multiple cleaning mechanisms.

• Aqueous systems do not pose air-emission or environmental risks.

• Formulations can be sprayed safely at higher temperatures, thereby allowing for higher levels of two key cleaning parameters: mechanical action and temperature.

• Aqueous systems are often more cost-effective for the overall process.

• Disadvantages of aqueous-formulated detergent cleaning

• It may be necessary to install spray devices in vessels and modify associated equipment to achieve the liquid coverage across surfaces that is essential for cleaning.

• Cleaning agents are often proprietary formulations and only available from a limited number of suppliers.

• Residue limits for the detergent must be set and validated analytical methods are required for detergent residue analysis.

• Selecting appropriate cleaning agents may need laboratory cleaning studies or vendor recommendation.

Numerous API manufacturers have weighed these advantages and disadvantages, and have adopted aqueous detergent cleaning processes where possible for API manufacturing. However, successfully converting from solvent cleaning to aqueous cleaning is not simple. It requires an understanding of the chemistry, equipment, engineering, validation and cost issues associated with solvent substitution, and a process to systematically address these issues.

Solvent substitution issues

Selecting the right chemistry. The solubility of an API in an organic solvent is a function of the polarity of the solvent. In the case of aqueous cleaning, solubility is a strong function of the pH of the cleaning solution. Certain functional groups are more soluble in an alkaline cleaning solution and others are more soluble in an acidic cleaning solution. For example, amino acids, diols, triols, organic acids, polysaccharides, saturated oils and waxes are generally cleaned better with an alkaline cleaner. Aldehydes, alkaloids, amines, bicarbonates, carbonates, ethers, insoluble hydroxides, ketones, metal oxides, pyridines and pyrrolidines are more likely to be cleaned with an acidic cleaner. However, selecting cleaning agents on the basis of solubility alone may be inappropriate because other mechanisms may be involved in the cleaning process. Table I lists the solubility data of APIs in detergent systems. These detergent chemistries were evaluated in the laboratory and the cleaning processes were successfully implemented in the field process equipment. For confidentiality reasons, the exact drug active and cleaning chemistries are not disclosed. Three detergents were used: a potassium hydroxide-based detergent (D1); a glycolic acid-based detergent (D2); and a phosphoric acid-based detergent (D3).

Studies. Cleaning studies using stainless steel coupons were performed. Approximately 10 g of each drug active was coated onto an area of approximately 100 cm2, allowed to dry overnight and subjected to agitated immersion cleaning. These studies showed that drug actives A1 and A3 could be cleaned from surfaces using a 2% solution of detergents D1 and D2, respectively, at 60 °C for 45 min under agitated immersion conditions.

In the case of drug active A2, cleaning studies showed that it could not be cleaned off coupons by any detergent (D1, D2 or D3) alone. However, detergent D3 in combination with another detergent additive was able to clean active A2 at 60 °C with 90 min of agitated immersion. This demonstrates that the solubility of an API in a cleaning agent may be misleading and may not be a good measure of the agent's ability to effectively clean that residue from surfaces. Although the solubility of these APIs in the detergents was very low, they were subsequently cleaned successfully from process vessels using aqueous detergents. These provided cleaning mechanisms other than solubility, including solubilization, emulsification, wetting and dispersion.

Selecting the right chemistry and cleaning parameters is very important for the success of a solvent substitution process. Cleaning studies can be conducted in the laboratory by coating the API residue on surfaces, drying or baking them to simulate actual process conditions, and then screening various detergents to determine the right combination of chemistry and cleaning parameters. These parameters include the cleaning time, agitation levels, the cleaning agent concentration and the application temperature. The studies should account for worst-case conditions that may be encountered in the process equipment. Among the issues to consider are the soil residue condition (dried, baked and caked), the nature of the surface, surface finish, ratio of the soil to surface area, soil redeposition and foaming.3

Figure 1: Concentration of active and cleaning agent.

Equipment design issues. The API industry uses a variety of substrates for manufacturing vessels. The most common of these are 304 or 316 stainless steel, glass-lined, PTFE-lined and hastelloy substrates. Evaluating the compatibility of cleaning agents with these substrates is important to ensure product integrity and equipment protection.

In the case of glass-lined reactors, for example, the concentration and temperature of alkaline cleaning solutions should be limited because of the possibility of etching glass surfaces by overuse of alkaline solutions. Data and technical support from cleaning agent and equipment vendors can be used to determine the isocorrosion curves for cleaning agents on glass-lined equipment. Also, after using the alkaline cleaning agent, the glass surfaces must be thoroughly rinsed with water, ensuring complete coverage across all surfaces, to avoid localized etching of the glass when the reactors are heated. If data are not available, laboratory substrate compatibility studies should be conducted under typical exposure parameters and conditions.

Cleaning process pipes can be a major challenge in API manufacturing processes. One of the commonly encountered causes of cleaning problems in process pipes is the inadequate flow rate of the cleaning solution. It is recommended that the average flow rate of cleaning solutions through pipes is >5 ft/s. This is a commonly accepted design practice for clean-in-place (CIP) systems in the pharmaceutical and biotech industries. In an API manufacturing plant, it is not uncommon to find larger-diameter pipes in which this high velocity is not feasible. In those situations, a close examination of the pipe layout is necessary to assess the flow coverage, the potential for gas and particle entrapment and the level of flow turbulence.4 In large diameter pipes, such as overhead risers that lead to condensers, it may be necessary to install spray devices to ensure adequate coverage.

The proper design of process piping is essential for good cleaning performance; if they are not appropriately designed for cleaning, it is important to identify the problematic areas and then either modify the piping or develop an appropriate cleaning procedure. Dead legs should be minimized to have a length-to-diameter (L/D) ratio <1.5. In existing systems where this ratio is exceeded, the flow and coverage in those areas must be examined and the cleaning ability validated. This may involve one or more corrective measures, such as increasing the flow rate; reversing the flow direction if appropriate; changing the orientation of the dead leg; reducing the dead leg by pipe modification; diverting the cleaning solution through the dead leg as another loop; or dismantling and manually cleaning those areas.

Table II: API and cleaning agent concentrations and Table III: Ract and Rca values for each rinse.

Ball valves, which are used extensively in API manufacturing, can be difficult to clean. These may have to be manually cleaned if adequate CIP methods cannot be demonstrated and validated.

Cleaning strategies. In some processes, eliminating cleaning solvent may involve significant equipment modifications and the installation of several spray devices to provide complete coverage. If this is not feasible, an alternate approach could be to reduce, rather than eliminate, the use of organic solvents. This strategy would involve aqueous cleaning by agitated immersion to remove most of the residues from process equipment, followed by a solvent-vapour reflux step to reach some of the overhead areas that the aqueous cleaner cannot cover.

A follow-up solvent flush can be used for API residue-sampling purposes after aqueous cleaning, even in situations with established good aqueous cleaning coverage. Another reason for using a solvent after aqueous cleaning is to remove the residual rinse water from the system. The solvent used in this final step is usually polar, has good water miscibility, and is often the process solvent used in the subsequent batch.

Sometimes a solvent is used to remove or recover the gross product residue from the processing equipment before using an aqueous detergent. In this case, the process solvent of the batch that was just produced is used.

Cleaning validation. General guidelines for cleaning validation are discussed in section 12.7 of the ICH Q7A GMP Guidance for APIs. Industry practice and sample calculations for establishing acceptance criteria of actives in API manufacturing have been discussed in the literature.5 Acceptance criteria for a cleaning detergent should be established using a similar approach based on its toxicity and effects on any subsequently manufactured API, as well as how those APIs are used in finished drug products. Cleaning validation for detergent removal must also be done when detergent cleaning is performed, whether it is between batches of the same API or different APIs.

Total organic carbon (TOC) analysis is commonly used as an analytical method for detergent residues in the finished pharmaceutical and biotech industries. However, as a non-specific method, it is not very commonly used in the API industry because of the potential interference from background organic carbon contained in the solvents used for manufacturing and cleaning. In addition, the water used for cleaning in an API facility may not always be low-TOC water.

Other analytical methods that are used for detergent residues may include ion chromatography, atomic absorption, conductivity, titration, high-performance liquid chromatography (HPLC) and UV spectrophotometry. HPLC is a commonly used method for the analysis of drug actives and for that reason is frequently chosen for cleaning agent residues as well.

Case study

An API manufacturer was using large quantities of methanol to clean drug actives from its manufacturing process trains. The original cleaning procedure used 9–11 methanol batch refluxes to meet the acceptance criteria of some water-insoluble APIs in a particular process train.

The company decided to evaluate aqueous cleaning procedures to reduce excessive solvent use. Initially, the aim was to have the same cleaning procedure for all the actives manufactured in the process train and to achieve methanol reduction without the need for equipment modification. Laboratory cleaning studies were conducted using aqueous cleaning chemistries.6 Based on the results of the studies, the company converted and validated its cleaning process to use a combination of aqueous cleaning followed by methanol cleaning. A process wash with 2% alkaline detergent (D1) followed by 2% acidic detergent (D2), both at 70 °C, was shown to significantly reduce the residue levels by an agitated immersion process.

The cleaning agents (D1 followed by D2) were not required for all residues, but were used so that a single cleaning process could be employed for product residues that required either an alkaline or an acidic approach. This general procedure helped with grouping strategies and with standardized cleaning procedures.

Because this aqueous agitated immersion process did not give complete coverage across all surfaces (particularly the overhead risers and condensers), and consequently did not achieve complete residue removal, two methanol flushes were needed after the aqueous cleaning steps were completed. Nevertheless, this new process was implemented and validated because it significantly reduced the amount of solvent used and provided a standardized cleaning procedure.

A few years later, driven by the Clean Air Act regulations and increased pressure to improve capacity utilization, the company decided to conduct a cleaning trial in the manufacturing facility to evaluate the possibility of further reducing cleaning time and a further reduction or elimination of methanol.

Cleaning trial. For the field trial, a section of the process train was used. This consisted of a 1000 gal glass-lined reactor; an overhead riser leading to a condenser; a filtration system at the discharge of the reactor; and a process pump and associated lines.

Laboratory evaluations showed that the drug active used for the trial had a water solubility of only 0.025% at room temperature. They also indicated that cleaning could be done at lower temperatures if higher concentrations of the detergent were used. It was determined that using ambient temperature saved time. Not heating the cleaning solution would more than offset the additional cost of the increased cleaning agent concentration that would be required.

The trial was conducted using a 4% solution of detergent D1 at ambient temperature for 30 min, based on supporting laboratory cleaning data. The 6 in. riser and 2 in. condensate return lines were cleaned by recirculating the cleaning solution. The highest point in the riser near the condenser had a nozzle. This was used to bleed off any air that was trapped in the 6 in. pipe to allow complete liquid coverage in the large-diameter pipe. The bleeding of the cleaning solution was done by inserting a flexible tube into the reactor.

The process was segregated into three loops to ensure coverage of all the surfaces. Loop A consisted of the glass-lined reactor, pump, filter units and pipes, loop B the reactor, pump, vent line and vent return line, and loop C the reactor, pump, riser, condenser and the associated pipes. The centrifugal pump at the discharge of the reactor had a 15 hp motor with 3600 rpm, and delivered approximately 75 gal/min. The pressure at the pump discharge was 35–40 psig, depending on the loop.

The aqueous cleaning process consisted of the following steps:

• Each of the three loops was prerinsed once to drain with deionized water for approximately 5 min.

• The reactor was filled with 400 gal of 4%-by-volume cleaning agent in deionized water. The solution was recirculated through each of the three loops (A, B and C, in that sequence) for 34–41 min and finally discharged to drain.

• Four hundred gallons of deionized water was recirculated through the three loops in sequence and then discharged. The recirculation time for each of the three loops was 6–8 min. The process was repeated four times.

Samples were drawn during the wash and water-rinse steps and analysed for active ingredient using UV spectrophotometer analysis and, for the cleaning agent, by conductivity.

Observations and results. The concentrations of the API and the cleaning agent are tabulated in Table II and illustrated in Figure 1. After the first rinse, the reduction in the cleaning agent concentration between Rinse 1 and Rinse 2 was not as high as expected. For any rinse, the ratio of actives (Ract) is defined as the concentration of drug active in the preceding rinse (or wash) divided by the concentration of active in the present rinse. In a similar manner, the ratio of cleaning agents (Rca) is defined to be equal to the concentration of cleaning agent in the preceding rinse (or wash) divided by the concentration of cleaning agent in the present rinse. The values of Ract and Rca, calculated from the data in Table II, are tabulated in Table III for each rinse. For Rinse 1, Ract has been estimated from the volume-weighted average of actives in the wash solution for the three loops.

The wash step and each of the subsequent four water rinses in this trial used the same amount of water (400 gal). Also, earlier studies using the cleaning agent D1 showed that it is very water soluble and can be easily rinsed; thus, almost all of the cleaning agent should be present in the rinse water and very little adhered to the equipment surfaces after the first rinse.

If it is assumed that almost the entire amount of cleaning agent residue was in the rinse water, then it can be concluded that there was liquid held up in the system approximately equal to the rinse volume (400 gal) divided by Rca. Therefore, for Rinse 2, for example, that value is 400 divided by 11.4, or 35 gal. This means that approximately 35 gal of the rinse water was not drained out of the system at the end of Rinse 1.

Because a holdup of 35 gal of rinse water was much larger than would be expected for this system, the lines were opened after Rinse 3 to investigate the cause of the low value of Rca. A significant quantity of rinse water was discovered and drained from the flexible hose that connected the two filter housings. It was estimated that the quantity of drained water was 20–25 gal. After the draining, Rinse 4 was done. As Table III shows, the Rca values increased significantly to 39. This implies that the holdup volume in the entire system after appropriate draining was now only 10.3 gal (400 divided by 39).

Rca can be viewed as a rinsability factor. Higher Rca values for the rinses are desired. For Rinse 1, the Rca value was low because the cleaning agent drained during the wash cycle was not yet rinsed off the surface. The Rca values for subsequent rinses depend on how freely rinsable the cleaning agent is and on the configuration of the process equipment and resultant holdup volume after rinsing and draining. This shows the significance of the need to design process vessels and pipes that are easily drainable. Drainability plays a significant role in preventing cross-contamination and is important for improving cleaning process efficiency and conserving rinse water.

The rinsability factor is related to the slope of the curve in Figure 1, where the y axis is on a logarithmic scale. A process with good liquid coverage using a freely rinsable component, such as the cleaning agent in this case, will exhibit a straight line. A steeper slope implies more drainable process equipment, a higher Rca value and, consequently, a lower rinse-water volume requirement.

By accounting for the drainage of the wash solution and applying an Rca value of 39 for the subsequent rinses (rinse water drained off without holdup, as in Rinse 4), it can be calculated that the cleaning agent concentration in the third rinse would have been less than 1 ppm. If the problem of the holdup in the flexible hose had been detected and proper drainage restored early, the fourth rinse could have been eliminated.

Table II also shows that the Ract values are approximately equal to the Rca values for Rinse 1. This suggests that most of the active ingredient and cleaning agent remaining in the system after the wash step was in the draining solution. However, as the rinsing process continued, and as the concentration of the actives dropped, the Ract values decreased significantly while the Rca values remained high. This suggests that the contribution of soil coming off system surfaces during the rinsing process became more significant. That is, either the soil was not as easily rinsed off the surface as the cleaning agent was or there was a location where cleaning was poor. This is also confirmed by the fact that for Rinse 4, Ract was 1 (because Rinse 4 had higher levels of active than Rinse 3).

To investigate this, the filter housing, which was the suspected cause, was opened and examined. Some visible bulk active residue, which was caused by inadequate coverage as a result of air trapped in the system, was found. The area was manually cleaned and appropriate measures suggested to bleed off the air and allow for good coverage in that area for future cleaning.

Other areas of the process were closely examined at numerous locations, including the dome of the reactor and pipe nozzles. All the areas were visually clean.

After aqueous cleaning, a reflux and flush was conducted using 500 gal of methanol through the three loops. The rinse solvent was analysed and found to contain 6 ppm actives in the methanol. The acceptance criterion was 10 ppm for this active.

Further improvements in the aqueous cleaning procedure were made following the trial, and the new aqueous-based cleaning process was implemented at the manufacturing plant.

Conclusion

This example demonstrates the feasibility of reducing or eliminating cleaning solvent use. It highlights some of the issues related to solvent substitution and shows that close monitoring of the wash and rinse solutions during a cleaning trial and appropriate analysis of the data can provide invaluable information to improve the cleaning process.

Solvent substitution and the use of aqueous cleaning for API manufacturing processes has been driven by a number of factors such as the relatively high cost of solvent acquisition, storage and disposal; increasing environmental and health and safety regulatory pressures; inefficiency and often ineffectiveness of the solvent-based processes; and overall economics. Efficient and successful conversion from solvent-based to aqueous-based cleaning is feasible with appropriate investment in equipment modifications and attention to the details of cleaning process design and validation.

References

1. R.J. Romanach et al., "Combining Efforts to Clean Equipment in Active Pharmaceutical Ingredient Facilities," Pharm. Technol. 23(1), 46-58 (1999).

2. S. Mohan and A. Shimada, "Navigating Through the Pharmaceutical MACT Maze," Chem. Processing 64(1), 44-47 (2001).

3. G. Verghese and J. Thomas, "Cleaning Agents for Biopharma Manufacturing," Genetic Engineering News 23(6), 46-52 (2003).

4. G. Verghese, "Developing a Validatable Cleaning Process," Proceedings of the 1999 Interphex Conference (Reed Exhibition Companies, Norwalk, Connecticut, USA, 1999) pp 461-469.

5. D.A. LeBlanc, "Establishing Scientifically Justified Acceptance Criteria for the Cleaning Validation of APIs," Pharm. Technol. 24(10), 160-168 (2000).

6. STERIS Corporation, Mentor, Ohio, USA, literature # 410-500-3611 (1999) www.steris.com