Risk-Based Cleaning in Biopharmaceutical API Manufacturing

Cleaning Validation (CV) is driven by regulatory expectations to ensure that residues from one product will not carry over and cross contaminate the next product. Regulatory scrutiny is more rigorous in a multi product facility compared to a single product facility. Companies are usually cited either for not having a sound cleaning validation or not meeting the protocol acceptance criteria. Because failing a protocol acceptance criteria is considered a substantial regulatory risk, companies are forced to spend money and resources even though there is minimal or no product risk.

It is vital for a successful cleaning validation to have appropriate acceptance criteria. In developing the acceptance criteria, companies may adopt a conservative approach either to prove that they have a sound cleaning validation program or to ensure that field data (results) will reflect the acceptance criteria. The Food and Drug Administration’s (FDA) guidance for determining residue limits is that they must be logical (based on understanding of the process), practical, achievable and verifiable. In validation, adequacy of each cleaning procedure requires demonstration that it can reliably and effectively remove or reduce residues to an acceptable level such that residues from the production of one product will not carry over in significant amounts to the next product. Companies today are faced with the challenge of reducing validation costs in an environment that demands increased compliance with current good manufacturing practices (cGMP). FDA’s initiative in pharmaceutical cGMPs for the 21^st Century is a science and risk0-based approach to product quality regulation. The risk-based approach will enhance the ability to focus on identifying and controlling critical factors that effect process and product quality.

When setting the specification for a production process it must reflect product quality so appropriate action can be taken when it deviates from specification. The purpose of setting process input limits is to consistently produce a quality product. On the other hand, the purpose of acceptance criteria in cleaning validation is to ensure acceptable levels of carryover residue. If the same acceptance criteria are used for cleaning validation in cell culture, purification, and formulation/filling processes, a good cleaning validation program can be penalized because the acceptance criteria are too rigid. Therefore, a cleaning validation program based on risk assessment with justifiable and achievable acceptance criteria is crucial.

Many biopharmaceutical products are proteinaceous, and their manufacturing involves media and buffer preparation, cell growth, cell harvest and processing, product purification, and other steps. Biopharmaceutical residues are often composed of proteins, lipids, simple and complex sugars, and salts. For biologically derived products, contaminants could be any one or more of the various organic and/or inorganic residues introduced or derived during the production process. Many biopharmaceutical processes also require decontamination with steam prior to cleaning; heat denatured residues are a common challenge. Proteinaceous residue or polysaccharide removal can be difficult due to denaturation, insolubility, foaming, and other factors. The removal of lipid, simple sugar and salt residues are often less difficult. Validation of a cleaning process requires the demonstration that the cleaning process can reliably and effectively remove or reduce residues to an acceptable level. The validation of a multi product cleaning process is more challenging compared to a single product.

Worst case validation testing strategies reduce the total quantity of validation studies for a system or process. By demonstrating that the cleaning process is consistent and effective for the worst case conditions, all other conditions are, by default, assured of consistent, effective operation. Worst case may include highest soil (product and others) concentrations in the hardest to clean area. Due to a large number of variable parameters that exist for the entire cleaning process, the worst case cleaning concept is applied to unit operations or equipment matrices.

In a traditional approach, acceptance criteria are derived from industry standards, past experience, company policy and capability study data. Some companies choose to perform development studies to determine achievable cleaning performance and to set the acceptance criteria based on study results. A clean-in-place (CIP) cycle is developed for worst case cleaning scenario in terms of soils, dirty hold time and equipment configuration followed by development runs. If they meet the acceptance criteria, then a validation protocol is developed and a minimum of three runs are performed. When all validation runs meet acceptance criteria, a summary report is written and approved, and the cleaning process is considered validated. However, the company faces a dilemma when validation runs fail. Root cause analysis may show that acceptance criteria are too rigid and are not scientifically based. Because revising acceptance criteria may not be a viable option when a protocol fails at this stage, the company is forced to modify the cycle and return to the development stage.

In a risk based approach, the CIP will determine acceptable carryover and ensure no product impact. Acceptance cleaning will depend to a given extent on what other products are manufactured with the same equipment. The level of any residues from earlier cleaning processes present in the final Active Pharmaceutical Ingredients (API) must be evaluated in the light of the effect on the manufacturing process for the API and on the finished drug product. Product, soils and equipment are reviewed to determine the Maximum Allowable Carry Over (MACO). Acceptable criteria are determined by dividing the MACO by a safety factor. The safety factor is selected from toxicology and reliability data, the cleaning agent used and company policy. The stage in API production should be considered when establishing acceptance criteria. Where residues from upstream cleaning steps will be reduced or removed in the subsequent processing steps should be taken into account when setting the safety factor. For example, if the maximum allowable level of specific residue in the finished drug product is 5ppm, and the maximum allowable level in the API constitutes 0.1 percent of the finished drug product, then the maximum allowable level in the API is calculated to be 5,000 ppm. The effect of the process/product residues on the manufacturing process should be considered when setting acceptance criteria. Where process residue may be allowed in large concentrations based on carryover into the final product, the effects of the residue on the manufacturing process should also be evaluated. A proteinaceous residue left in large amounts on the surfaces of a bioreactor could provide a substrate and energy source for the growth of undesirable organisms.

CIP cycles are developed for worst case cleaning scenarios in terms of soil type, dirty holding time and equipment configuration. In choosing cleaning solutions, cost, removal of cleaning chemicals, toxicology (product impact) and environmental (waste removal) are considered. Development runs should be conducted to test the cycle prior to initiating CIP validation. These runs should be designed to evaluate CIP cycle parameters, cleaning chemistries and equipment configurations. They should demonstrate a proven acceptable range for the cleaning process. Determining a point of failure for the cleaning process is not required. The developed process should be sufficiently robust to achieve a high degrees of assurance of reducing residues to an acceptable level. When the results of the development runs meet the acceptance criteria, a validation protocol is developed and the required number of runs (typically three consecutive runs) are performed. When all runs met the acceptance criteria, a summary report is approved and CIP is considered validated. If acceptance criteria fail, an investigation is performed to determine the root cause of the failure. In this investigation, the company focuses its resources on the operational and cycle failure, not acceptance criteria. For equipment failure, the impact on the run is evaluated and the run is accepted or repeated. However, when a validation run fails and analysis reveals that the cycle is not effective, then the cycle will be modified for better cleaning performance. There will be no change in acceptance criteria. Effective cycle development will significantly increase the probability of first pass success in validation runs.

Critical parameters for CIP cycles include cleaning detergent concentration, duration of the cleaning and rinse steps, flow rates, pressure and temperature of the cleaning and rinse steps. Critical cleaning parameters for manual cleaning and clean-out-of-place (COP) include temperature of cleaning and rinse solutions, duration of cleaning and rinse steps, and cleaning detergent concentration.

For risk based acceptance criteria, the following factors should be considered during the development of acceptance criteria; Maximum allowable carry over (MACO) and safety factors. Process risk vs. patient risk. Manufacturing stage (pre, post, during purification). Cross contamination between products or product intermediates. Single vial concept and worst case cleaning.

The MACO approach can be used for API, cleaning chemicals, and other soils. The MACO principal has been discussed in detail and calculations are shown in various literature. For biotechnology processes, detection of the desired API can be challenging, let alone detection and quantitation of a specific contaminant. When Steam-in-place is used for decontamination prior to CIP, API protein would be denatured. No cross contamination for API is expected for a single product, hence MACO for active ingredients may not apply. In that case, it is expected to establish MACO for other soils and denatured API in terms of total organic carbon (TOC) levels. For cleaning chemicals, the lethal dose (LD50) or the no observed effect level (NOEL) can be applied. The safety factor will depend on the data type. A safety factor of 1000 may derive from 10 for intraspecies, 10 for interspecies and 10 for route differences.

Process risks are typically risks to the yields or impurity profiles of the API during the manufacturing process. Process risks are generally categorized as a producer risk as they may affect the cost of manufacture in case of yields, or make the product unsuitable for use, or require re-processing  in the case of atypical impurity profiles. These risks also present a regulatory risk as they may indicate a manufacturing process not in a state of control. These risks are usually less severe than patient risk and may require a smaller safety factor than patient risks. Patient risks are direct or indirect risks of adverse events due to ineffective cleaning. One of the most severe risks is cross-contamination between two different products. This is especially true for potent or cyotoxic products. Where possible, risks of cross-contamination between products should be minimized through various mistake-proofing or mitigation methods such as use of dedicated or disposable equipment, on-line measurement of residue levels during cleaning, and other viable methods.

A significant consideration in assessing patient risk is the manufacturing stage in which a contaminant is introduced. A contaminant introduced in the cell culture/fermentation step will be less likely to be present in the final product than contaminant introduced post purification. The number of dilution or purification steps should be factored in when assigning the risk levels. Process risk should also take into account the process step in which the contaminant is introduced and the effect of the contaminant on subsequent processing steps. Process risks such as membrane or column, atypical elution profiles, etc. should all be factored into account when assigning risk based on the manufacturing stage

Cross-contamination is a major concern addressed by the cleaning process. When designing a cleaning validation program and evaluating process/patient risks, the potential for cross contamination between products should be assessed. For biopharmaceutical products, specialized consideration should be given in areas where there is a concurrent manufacture of multiple products manufactured by the same equipment. A useful tool for evaluating routes for cross contamination between products is a fault tree analysis (FTA), where a team assembles a logical diagram that illustrates failures required for a contamination to occur.

The single vial concept is a conservative approach to cleaning validation in which it is assumed that all of  a given contaminant resides in a single vial of product. This approach is frequently used in pharmaceutical tablet manufacturing facilities, where there is a significant risk of a contaminant ending up in a single tablet due to batch non-uniformity. This approach leads to stringent acceptance criteria as the single worst-case swab result is used for acceptance or rejection of an entire validation study, even though this worst-case location may only represent a fraction of the total area cleaned. In biopharmaceuticals that are aseptically filled, this approach may be used for any contaminants that occur after the final sterile filtration. This approach, while appropriate for final formulation and filling, is not appropriate for APIs that are further processed (diluted and mixed) and sterile filtered prior to final filling. The processing and filtration steps make it much more likely that contaminants will be uniformly distributed throughout a batch rather than in a small volume of the product. Averaging swab values may be more representative of actual process risk for APIs.

Worst case cleaning is cleaning at a set of parameters within the proven acceptable range for the cleaning process that represent the least effective allowable cleaning process. Worst case parameters should be determined during cycle development, with the exception of hold times between the end of use of the equipment/system and the cleaning of the equipment (dirty hold times). The cleaning parameters should run at nominal values during cleaning validation studies for APIs. The performance of equipment under worst case conditions should be evaluated as part of the cycle development studies for cleaning. Validating dirty hold time greater than the maximum dirty hold time is not recommended when an extended hold may change the nature of soils(e.g. flaking). Because dirty hold times between equipment use and cleaning can significantly affect the effectiveness of the cleaning process, worst case hold times should always be evaluated during the cleaning process validation studies. For API manufacture, the amount of safety factor over the maximum allowable dirty hold time may be varied based on the patient and product risk posed by the contaminants.

To validate cleaning of multiple pieces of equipment, a matrix approach can be employed. Equipment of similar size and configuration exposed to similar soils can be grouped, and representatives of the group can be analyzed after cleaning operations. For CIP, a single member of the group is evaluated over three runs while other members are analyzed in a single run. In some cases, three runs in a representative tank in worst case soiling conditions and no run in some tanks may be justified through development studies and monitoring. For manual cleaning and COP, a single member can be analyzed over three runs and other members may not require any testing. The cleaning validation protocol will define the testing requirements for a group of like equipment and justify the test plan. Part of the justification for using a matrix approach should be a risk assessment of the cross contamination effect between products in a given family. Items that are high risk due to the difficulty of cleaning or low MACO should be used as worst case or qualified separately .

To apply the matrix approach, the manufacturing process is reviewed for three main parameters; equipment type (size, surface finish and configuration), cleaning process and soil characteristics. Soils are considered to be any residue that could possibly remain on the equipment after the manufacturing process. All manufacturing equipment in the process segment is inventoried and grouped by cleaning procedure and equipment type. The equipment is considered to be of the same group if it is cleaned by the same cleaning procedure and is of equivalent size shape and construction. In cases where equipment is exposed to multiple soils in normal usage, the most difficult soil to clean or the highest risk soil is identified. This soil is used as the representative soil for cleaning validation. Representative soils are selected based on solubility, reactivity, toxicity and difficulty in cleaning.

The various risk analysis tools recommended for pharmaceuticals include FTA, failure mode and effect analysis (FMEA) and hazard analysis and critical point control (HACCP). ISO 14971:2000 is a risk assessment for medical devices that requires one to define usage, identify hazard and estimate risk. FMEA can be applied to identify hazard and occurrence probability. It is a preventative tool, with a bottom up approach to identify all potential failures of a product, process or system prior to use as well as assessing the effects or consequences of identified failure modes. FTA is a structured top down approach used to identify the factors that contribute to a failure. It can be used by itself as a risk analysis method or in conjunction with FEMA to identify failure risks.

Cleaning validation demonstrates the removal of the production process residues and cleaning agents. A key question is relating what is measured in the rinse water routine monitoring to possible contamination levels of that residue in the subsequent product. Sampling should be indicative of the contaminants and avoid redundancy. Swabbing and visual inspection are utilized to directly inspect product contact surfaces and is the primary method for assessment of surface cleanliness. Indirect testing methods should be used for inaccessible area (e.g. plug port). Potential critical sites or areas where residues are likely to accumulate should be identified for swab sampling. Rinstate sample testing for pH, conductivity and total organic carbon (TOC) are direct testing methods to confirm direct inspection results. Rinstate pH and conductivity data may provide effectiveness of the cleaning agent removal. The TOC assay will detect organic carbon residues from product residues, cell culture/fermentation media and other organic materials including organic components of formulated cleaning agents. Control of the microbial and pyrogen contaminations should be monitored through bioburden and endotoxin testing, respectively.

Visually clean is a regulatory expectation for all cleaning processes. Visual analysis is used to evaluate the effectiveness of cleaning procedures and cycles by detecting the presence of visible residues. Residual materials can usually be detected down to the 1 to 4 mg/cm level. Sensitivity will depend on soil type and concentration and operator’s qualification. Sensitivity could be determined during development runs using actual equipment, soil and operators as part of a simulation study. When a study uses actual equipment, soil and qualified operator, it may be justified that no swab sample is required for validation on the basis of a relationship between the swab and rinse results.

Due to acidic or caustic properties of the cleaning agents, it is not expected to build up microbial contaminants during cleaning. However, the distribution of the contaminants and proliferation of the microbe during clean hold could be a concern. Validation should identify the source of the contamination and address them. Microbial analysis is used to evaluate the effectiveness of the cleaning process for reducing the bioburden level. When SIP is used post CIP prior to equipment use, a substantial increase in bioburden load may affect SIP performance. The acceptable limit for bioburden is equal to or lower than the product specification for each manufacturing process segment. The total coliform specification is 0 CFU/mL. Endotoxin analysis (Limulus Amoebocyte Lysate Assay) is used to evaluate the effectiveness of the cleaning process for the reduction of endotoxin. In the cell culture and purification process, the acceptance limit for endotoxin is equal to or lower than the process specification.

pH and conductivity indicate rinstate acidity or alkalinity. They can be used as indicators for the removal of cleaning agents. If the product is acidic or basic, rinstate pH and conductivity values may be used to evaluate the presence of product residues in the rinse. pH analysis is used to evaluate the effectiveness of the rinse procedures by detecting trace quantities of basic or acidic compounds used in the cleaning process. pH analysis of the final rinse water may be especially appropriate for process steps such as column chromatography or tangential flow filtration, where pH adjusting agents such as sodium hydroxide (NaOH) are also used as cleaning agents. If traces of NaOH are present after cleaning, there would be minimal risk as it is already a process component. Measurement of the final rinse water pH can be a sensitive measure of residual cleaning agent in this case. Concentration curves correlating the concentration of NaOH in the final rinse water pH should be determined to set the acceptance criteria for final rinse water pH. Conductivity is used to evaluate the effectiveness of rinse procedures by detecting trace quantities of cleaning chemicals during water-for-injection (WFI) rinse. A relationship between conductivity and concentration of cleaning chemicals needs to be established. Acceptable carry over of cleaning chemicals should be based on its toxicity such as LD50 and NOEL. A typical CIP process consists of an initial purified water rinse to drain, followed by a caustic step, and acid step, and the final WFI rinse. In this case, conductivity due to the presence of the acid solution is dominant. However, verification should be in place for complete removal of caustic.

Rinstate TOC is an indicator of the presence of carbonaceous material in the rinse. Although rinstate TOC is an indirect test, it can be a good indicator of the presence of product/buffer/media or cleaning chemicals in the rinstate. TOC assays for rinse water and surface swab samples are non-specific and are performed to detect organic residues such as residual product and other organic components. To set acceptable limits for these assays, three factors are considered; MACO, suppression of the bioburden/endotoxin during post cleaning storage, and CIP system capability. For Multi product facilities, TOC acceptance criteria should be established on the basis of MOAC. For single product manufacturing with dedicated equipment and no concern for cross contamination of active pharmaceutical ingredients (API) application of the MACO may not be applicable. Equipment must be cleaned to the point where residuals will not significantly support growth of microorganisms between equipment uses Equipment clean hold studies should support that equipment cleaned to TOC acceptance limits for both surface and rinse sampling meet bioburden and endotoxin acceptance limits after storage. Some equipment due to size and configuration is impossible to swab. In these cases, rinse sampling is the primary method of evaluation and protocols will include justification of the rinse sample methods utilized. Some buffers contain no organic materials and as a result, TOC testing is not an appropriate test on buffer tanks that contain these buffers.

Swab TOC acceptance criteria are usually derived from uniform distribution of the soils throughout the equipment stream. It is expected that the hard to clean worst location swab area will have a higher TOC compared to the easy to clean surfaces. Using the same acceptance criteria for hard to clean and other areas may be justified recognizing that it represents the worst case acceptance scenario and using separate acceptance criteria is not feasible. There is no need to assume uniform TOC concentration on the surface after CIP when actual surface TOC value after CIP validation run is available.

Biotechnology production processes may use chemicals of a significantly toxic nature such as Methotrexate or Cyanogen Bromide. The acceptable level of toxic chemicals in rinse water is established based on the NOEL drug toxicity rationale. Risk analysis for toxic chemicals should take into account not only patient risk, but risk to production employees and to the environment. Risk analysis should also consider the stage of the production process in which toxic chemicals are used. For example when Methotrexate, a selection agent, is introduced into the manufacturing process (e.g. spinner flasks) it is diluted into large bioreactors prior to purification. This dilution effectively reduces the concentration of Methotrexate to a negligible level in the API prior to purification. Therefore, Methotrexate should not be considered a residue indicative of the cleaning procedure effectiveness, except in the early stages of the production process (spinner flasks and initial scale-up).

System capability will demonstrate the robustness of the cleaning process. Cleaning validation studies should demonstrate that soil residues are reduced to below visual detection level (DL) by consistently meeting visual acceptance criteria. This demonstrates that the cleaning processes are capable of reducing TOC residues to visual DL on product contact surfaces. One way to determine system capability is by collecting samples at various times during the cleaning step. A plot of the rinse concentration vs. time will show a point where there are no significant changes in the rinse concentration with additional rinse time. A relationship between rinse concentration and surface swab results should be established. It is expected to perform a study to determine recovery of TOC residues from equipment surfaces. The linear regression line, passing through the average of negative control sample for each soil, can give a swab TOC concentration for a specific soil level

FDA defines Process Analytical Technology (PAT) as a system for designing, analyzing, and controlling manufacturing through timely measurements of critical quality and performance attributes of raw and in-process materials and processes with the goal of ensuring final product quality. PAT in cleaning can be applied to complement the validation and optimize equipment usage based on real time data. In both cases, it is expected to identify parameters that indicate equipment cleanliness. Usually this is done at the final rinse. A selection of parameters that can be monitored on-line is performed for the PAT application. Concentration vs. time plots for the parameters are determined with a safety factor. For indirect parameters, such as rinse concentration for soil concentration, relationships between surface soil levels and rinse water levels need to be established. These relationships should be used during validation studies to establish alert and action limits to be used with PAT. Types of PAT measurements that can be used to verify cleanliness after completion of the CIP cycle include conductivity/pH to measure residual cleaning agent, TOC to measure product or cleaning agent residuals, or UV absorbance for protein residues. Other types of measurements may be developed based on API and cleaning agent.

To avoid microbial proliferation, post-CIP drying, flushing with WFI, and SIP are utilized for biopharmaceutical processes. It is expected to demonstrate they maintain acceptable microbial quality between the time the equipment is used or steamed-in-place. For equipment not maintained as a closed system, recontamination with dust particles or airborne microbes can contribute to microbial proliferation when combined with moist surfaces. For equipment steamed-in-place, the validation of the SIP process must take into account the bioburden present in the equipment. Even when SIP factors bioburden into account, high microbial burden will cause increased endotoxin levels. Clean hold time is not a risk when SIP is performed immediately after CIP and equipment is maintained as a closed system between SIP and the next production cycle.