Destin A. LeBlanc

There are eight 'myths' surrounding cleaning validation for pharmaceutical process equipment and they revolve around issues associated with setting limits, selecting sampling procedures and using analytical methods to measure residues. These myths can be overly prescriptive or overly proscriptive in terms of what actions can actually be taken. Most add little value to a cleaning validation programme and can be ignored because they have no scientific or regulatory basis. I covered the first three of the myths listed below in Part I of this article and will discuss the remaining five this month.

1. Regulatory authorities do not like rinse sampling. 

2. You must correlate rinse sampling results with swab sampling results. 

3. You cannot use nonspecific analytical methods. 

4. If you use total organic carbon (TOC), you must correlate it with a specific method such as HPLC. 

5. Any measured residue is unacceptable. 

6. Dose-based calculations are unacceptable. 

7. Recovery percentages of different spiked-levels should be linear. 

This myth, which is related to myth 3, says you must correlate your TOC method with an HPLC method (or another specific analytical method). Again, this depends on what is meant by 'correlate'. If it means performing full method validation on both the HPLC and the TOC methods, the question should be: "Why is this necessary?" If a full method validation on the HPLC method is required, why not just use the HPLC method for measuring residues? 

After all, if I've gone that far with the HPLC method, it would be wise to utilize that HPLC method for measuring residues of the target residue because of the greater likelihood that I would meet my acceptance criteria (see myth 3 above for more on this). If correlation of TOC with an HPLC method just means performing my TOC method validation on known standards (perhaps standardized by an HPLC analytical method) to determine parameters such as accuracy and precision, then this is not really a special requirement. For example, I should do the same thing for the HPLC method. 

I can only speculate that this myth comes from someone who thought that TOC was unacceptable, and tried to add some additional supporting data to demonstrate that it was acceptable. However, as discussed in myth 3, TOC is (or should be) acceptable to regulatory authorities. 

If one is expecting correlation between TOC and HPLC methods on samples obtained in cleaning validation protocols, then this is an unreasonable expectation. Yes, the HPLC method will tell you how much of the target residue (for example, the active) is present. However, TOC is subject to a variety of 'interferences' (excipients and cleaning agent) that raise the measured carbon value. One cannot expect that the ratio of organic carbon from the active to the combined TOC is going to be constant in every sample (which could perhaps give a correlation). The bottom line is that TOC, appropriately validated, is sufficient without any correlation to a specific method. 

This is a really hard one to swallow. As analytical methods can measure residues at lower levels, it is inevitable that you may have a measurable amount of certain residues. If TOC is used as an analytical method, because of the various 'interfering' substances, it is also generally not true that all your samples will return to below the limit of detection.  

Before we discuss this, it is probably necessary to state that there are some conditions where any residue is unacceptable. In applying a visually clean standard, any residue related to the cleaning process (active, excipient, cleaning agent, degradant or cleaning by-product) that is visible on the surface should constitute a failure.

  Furthermore, for certain potent actives, allergenic actives or actives with reproductive hazards manufactured on nondedicated equipment, a reasonable standard is that the target residue is below the limit of detection by the best available analytical procedure.

  With at least those two exceptions, in most other cases measured residue is acceptable provided it is below the acceptance limit. 

The only basis for this myth is possibly the statements in both FDA and Pharmaceutical Inspection Cooperation Scheme (PIC/S) guidance that 'ideally' there should be no measured residues of detergent.

  FDA guidance modified this by stating that for 'ultrasensitive' analytical methods, 'very low' residues should be expected. I believe this myth may partly be because of the confusion of a 'residue' and a 'contaminant'. For the purpose of this article, a 'contaminant' is an unacceptable level of a residue.

This is much like the maxim in toxicology that 'the dose makes the poison'. The acceptability of some measured residues is support by FDA in its 

  In answer to the question: "Should equipment be cleaned to the best possible method of residue detection or quantification?" an emphatic "No" is the answer. In that same answer, FDA goes further to state that the real issues are whether the residue level is 'medically safe' and whether it 'affects product quality'.  

Of course, this assumes that you are setting residue limits appropriately (such that the residues are medically safe and do not affect product quality). It also assumes that you have appropriately validated the analytical method. One possible objection is that some of the methods used for calculating limits give very high limits. That objection will be discussed in myth 6. The bottom line is that measured residues must be at levels that are medically safe and do not affect the product quality. 

This relates to the fact that the traditional 0.001 dose calculation sometimes results in a limit that is extremely high. For example, in 2001 this traditional method of calculating limits for finished drug products was challenged by providing an example calculation where the allowable carryover to the next product was more than the next product batch size.

Unfortunately, in that paper the dose-based calculation was misapplied, in that the maximum dose of the next product was given as the maximum dose of the next drug active, not the maximum dose of the next drug product (which is the correct calculation for finished drug products). A correct calculation would have lowered the calculated limit by a factor of more than 1000.  

Furthermore, there are some restraints to the conventional dose-based calculation. One is that a 'default' limit of 10 ppm active in the next product is commonly used. That is, if the 0.001 dose calculation gives a value of greater than 10 ppm of the cleaned active in the next product, a default value of 10 ppm is used for subsequent calculations.  

This makes sense because if 24 ppm is the calculated safe level based on the dose calculation, then 10 ppm should also be safe. A second restraint is that no residue of the active (or any combination of cleaning residues) may remain, such that the equipment is visually dirty. Visually dirty can be subjective depending on such things as the nature of the residue, the distance and angle of viewing, and the lighting level.

  However, such criteria provide a further restraint on unusual situations where the dose-based calculation may allow what seems to be unreasonable levels of residues.  

This myth seems to have appeared and then disappeared as quickly. However, it should be clear from both FDA and PIC/S guidances that this approach is acceptable. FDA guidance does not explicitly say the dose-based calculation is acceptable, but it does footnote the 1993 article by scientists at Lilly who described this dose-based calculation.

  The PIC/S guidance is a little more explicit in that it calls for setting limits based on the most stringent of the 0.001 dose calculation, 10 ppm and visually clean.

  Ultimately, dose-based calculations are the current standard for most situations in which limits are calculated.  

This myth essentially takes what is a common requirement for analytical method validation (determination of a linear response over a certain range) and applies it to swab recovery studies. If three or more spiked amounts (concentrations) are used for recovery studies, sometimes analytical groups expect that the different amounts (concentrations) should exhibit a linear relationship. Now it is a reasonable assumption that, in general, the higher the spiked amount in a swab recovery, the lower the percentage recovery.  

However, for recovery studies it is not a reasonable expectation that the data will be linear in any sense. The reason being is that swabbing, like manual cleaning, is highly variable. One may spike at a fixed level, and one operator might get a recovery of 65% and another might get a recovery of 73%. There is not a significant difference between those two values (at least for swab recoveries). The variability in an individual swab recovery study is so wide that trying to establish linear relationships does not add any confidence or value. Now some might object that the difference between 65% recovery and 73% recovery is significant, in that it might be the difference between passing and failing in an executed cleaning validation protocol. The response to that concern is that if the difference between 65% and 73% is the difference between passing and failing, your cleaning process is not robust enough.  

My guess is that this myth originates from an analytical group that misapplied what is a reasonable requirement, that analytical methods are linear over a certain range, to different spiked levels in swab recovery studies, where linearity is not a reasonable expectation. This idea of linearity in recovery studies is probably best avoided by only doing recovery studies with (at most) two spiked levels. The first spiked level should be the acceptance limit. If a second spiked level is desired (it certainly isn't necessary), then it should be done at approximately 20% of the acceptance limit. Having two data points avoids the issues of determining linearity (or if you are perverse, it assures linearity). All in all: "Don't expect different spiked levels in swab recovery studies to be linear." There is just too much variability in any individual measurement because of the fact that swabbing is similar to manual cleaning. 

After talking about the variability of manual cleaning in myth 7, it might seem strange that I'm now talking about demonstrating the consistency of it. However, the issue in cleaning validation is not that the residue 

 have to be consistent (although that is something we would clearly like), but that the residue data have to be 

 the acceptance criterion. Because of the variability of manual cleaning, it is necessary to build a lot more robustness into the process by more adequate instructions in the written cleaning procedure, as well as more training and retraining of the cleaning operators. While the same level of control that is possible with automated cleaning is generally not possible with manual cleaning, a reasonable effort must be made to achieve more control than was generally achieved before 1990.  

This myth developed very early in cleaning validation, and was perhaps a response to regulatory authorities trying to justify why cleaning validation cannot and should not be done for manual cleaning processes. This appeal soon lost out and it is now a regulatory expectation to validate manual processes. FDA guidance says nothing directly about manual methods except for a comment relating to the need for "extensive documentation... and training" where operator performance is a problem.

  The PIC/S guidance states that "Manual methods should be reassessed at more frequent intervals than clean-in-place (CIP) systems."

  There may be some variations in interpretations of what this means, but it is clear that it is probably addressing the same issues of operator consistency that FDA guidance addresses (without specifying what the 'reassessment' might be).  

In any case, while there is definitely a trend toward automated cleaning systems (whether they are CIP systems or automated COP parts washers), it is expected that manual cleaning processes will be validated. Yes, they are more variable than automated systems. But, this just means paying more attention to detail in the standard operating procedure, and to more frequent training and retraining of operators to help assure consistency so that actual residues are consistently below the acceptance limits. The bottom line is: "You will validate manual cleaning processes." 

The eight myths covered in Part I and Part II of this article are not necessarily the only cleaning validation myths. Recognition of these myths, and their lack of both scientific justification and written regulatory justification, can help companies avoid unnecessary work, which adds little or no value to a cleaning validation programme. 

This paper is based on a presentation to the Capital Chapter of the PDA in Gaithersburg, MD, USA on 30 October 2002. 

www.fda.gov/ora/inspect_ref/igs/valid.html

, Cleaning Validation Technologies (March 2002).  

6. FDA, Human Drug cGMP Notes, 2nd Quarter 2001, PDA Letter 

7. A.L deMarco and S. Jansen-Varnum, "A Perspective on Equipment Cleaning — Industry Practice But Not cGMP", presented at Central Atlantic States Association of Food and Drug Officials Conference, King of Prussia, PA, USA (15 May 2001). 

Articles

In applying a visually clean standard, any residue related to the cleaning process that is visible on the surface should constitute a failure.

Dispelling Cleaning Validation Myths: Part II

Every regulated technology seems to come up with a list of what regulatory authorities supposedly say you should and should not do. Cleaning validation for pharmaceutical process manufacturing equipment is no different. Unfortunately, while many of these 'thou shalts' and 'thou shalt nots' have a partial basis in fact, they are actually distortions of the truth that come to have a life of their own. Hence I call them myths even though cleaning validation is only about 15 years old.  

This article will explore eight of these myths and attempt to explain the origin of each (although in many cases the explanation of the origin is just speculation on my part). In addition, I will try to explain why the myth is wrong, how something seemingly prohibited can be properly used, and how those things apparently required may be unnecessary. 

My list of myths is not intended to be exhaustive. The first three are examined in Part I of this article. Myths 4–8 will be covered in Part II to be published in a forthcoming issue. 

1. Regulatory authorities do not like rinse sampling.

2. You must correlate rinse sampling results with swab sampling results.

3. You cannot use nonspecific analytical methods.

4. If you use total organic carbon (TOC), you must correlate it with a specific method, such as HPLC.

6. Dose-based calculations are unacceptable.

7. Recovery percentages of different spiked levels should be linear.

The notion that regulatory authorities do not like or allow rinse sampling is false. FDA's cleaning validation guidance says: "There are two general types of sampling that have been found to be acceptable. The most desirable is the direct method of sampling the surface of the equipment. Another method is the use of rinse solutions."

  Some may want to emphasize that because direct sampling (i.e., swab sampling) is more desirable, it must follow that rinse sampling is less desirable. Although there may be certain logic to this, it overlooks the clear statement that both methods are acceptable. 

The Pharmaceutical Inspection Cooperation Scheme (PIC/S) guidance document says: "There are two methods of sampling that are considered to be acceptable, direct surface sampling (swab method) and indirect sampling (use of rinse solutions)."

  Again, this is a clear statement that rinse sampling is acceptable. I should point out that the PIC/S document goes on to say that a "combination of the two methods is generally the most desirable."  

Why, therefore, has the myth arisen that rinse sampling is unacceptable? In the early days of cleaning validation, rinse sampling was used inappropriately. For example, some companies using rinse sampling set limits such that the rinse sample was acceptable if it met compendia specifications. In other words, they worked by the maxim: "water-for-injection [WFI] in, WFI out, therefore, my equipment is clean." This use of rinse sampling is inappropriate, but still survives despite the fact that FDA's guidance document clearly states that "...it is not acceptable to simply test the rinse water for water quality (does it meet the compendia tests) rather than test it for potential contaminates [sic]."

  I should make it clear here that you can use TOC to measure a contaminant in the rinse water. However, the acceptance limit of TOC is not automatically 500 ppb: it must be justified based on traditional limit calculations, and may be higher or lower than 500 ppb. 

Rinse sampling has also been misused by not performing recovery studies. A concern of FDA (expressed in its guidance document) is the dirty pot analogy.

  Do you determine the pot is clean by evaluating the pot or the rinse water? One obvious answer is to test the pot. However, another is to test the rinse water providing it can be established that any residue on the pot would be present in the rinse water. Just as recovery studies for swab sampling are done by spiking model surfaces with the target residue and then sampling by the swab procedure, rinse sampling recoveries should also be performed by spiking model surfaces with the target residue and performing rinse sampling on those surfaces to demonstrate quantitative recovery.  

Lab rinse sampling recoveries cannot replicate production equipment rinsing. However, it is possible to simulate the rinsing conditions to demonstrate whether the rinsing process quantitatively removes surface residue. If it does demonstrate acceptable recovery, then the dirty pot analogy has been overcome. Where rinse sampling is used to demonstrate cleanliness of inaccessible surfaces (for example), rinse sampling recoveries should be performed to 'validate' that method. In this scenario, quantitative recovery is not 100% recovery. The acceptable recovery level is generally the same as that for swab sampling, which can vary from about 50–75%.  

I believe the misuse of rinse sampling has lead to the myth that its use is unacceptable. Correct use of rinse sampling includes 					

Carefully defining and controlling the rinse conditions.

Making sure what you analyse is a direct measure of the target residue.

Setting limits for that target residue (in the rinse solution) based on  scientific principles. 

This is the idea that to use rinse sampling, you have to correlate it with swab sampling results. If what you mean is: "I need to make sure I get passing results by both swab and rinse samples," there may be an element of truth in this. However, if what is intended is that there should be a direct 1:1 (or similar) mathematical relationship between swab and rinse sampling values, then it is unreasonable to expect this to occur. Why? Swab sampling and rinse sampling measure two different things.  

Swab sampling involves measuring the residue on a small area, which generally includes the worst-case locations (those most difficult to clean or likely to have unacceptable residue if cleaning is inadequate). However, rinse sampling covers a much larger surface area (perhaps the entire surface area of a manufacturing vessel), and, therefore, essentially averages the residue over all sampled surfaces. If failure occurs in swab sampling, it is reasonable to expect it to come from the worst-case locations and that, perhaps, other swabbed locations provide acceptable results.  

If such is the case, it may be possible (if not probable) that a rinse sample will give acceptable results. But, if rinse sample results are unacceptable, you can expect that at least one swab sampling site should have failing results. The assumption in this is that you have calculated your limits appropriately (and did not do something such as set rinse limits based on compendia specifications for water).  

I find it difficult to speculate on this myth's origins, except perhaps from an overzealous analytical group. Cleaning validation is hard enough in terms of ensuring necessary resources are available. Performing studies to mathematically 'correlate' swab and rinse sampling values does not add any value. What's more, do not expect them to mathematically correlate. 

It is amazing how this one myth, that nonspecific methods are either unacceptable or less acceptable than specific methods, persists. Specific methods measure the target analyte (usually a given compound) in the presence of expected interferences.

  Specific methods include HPLC developed for a given compound. 

Nonspecific methods measure a general property, but do not determine what compound that property is a result of. Methods include TOC and conductivity. TOC measures the organic carbon in a sample. In finished drug manufacture, the measured organic carbon in a cleaning validation swab sample may exist because of any combination of the active, excipient(s) and cleaning agent (as well as contributions for the blank, which could include the water, the swab and the vial).  

Somehow, nonspecific methods are viewed as less robust than specific methods. In fact, for cleaning validation, using a method such as TOC actually makes it more difficult for a manufacturer to meet its cleaning validation acceptance limits (again provided the limits are set correctly and the TOC data are converted appropriately into the target residue).

  If such methods were unacceptable, almost all biotechnology facilities would be shut down because TOC is used widely in the industry for measuring residues of the actives (in biotech, TOC is usually measuring degraded actives, but the measured TOC is expressed as if it were the undegraded active).  

The use of TOC is further supported by a Human Drug Current Good Manufacturing Practice (cGMP) Note from FDA in which it states: "We think TOC or TC can be an acceptable method for monitoring residues routinely and for cleaning validation."

  The cGMP note was replicated as a "Q&A for cGMP for Drugs" in 2002.

  FDA goes on to state the conditions that should be adhered to if TOC is used as the analytical method. However, the implication is that such methods are acceptable if used correctly.  

Where did the myth come from? My speculation is that it came from using TOC as an analytical method, but only setting limits based on compendia water specifications (that is, 500 ppb TOC). It should be clear from the Myth 1 discussion that in this case TOC is an unacceptable method (correctly stated, it is the limit setting that is unacceptable, but it is easy to see how this became "TOC is unacceptable"). This is further complicated by a statement in the PIC/S guidance document that analytical methods for measuring residues "should be specific for the substance assayed".

  Could this be interpreted that only a "specific analytical method" be used? Again, such an interpretation would wreak havoc with the biotech industry. If a specific analytical method was required, the statement would be more explicit. This statement is probably akin to that in FDA's guidance that for rinse samples, "a direct measurement of the residue or contaminant" should be made.

  Is TOC a direct measure of an organic active? I would argue that it is. This conundrum of what is meant in the PIC/S guidance should be recognized, but it should not deter us from using TOC appropriately for cleaning validation purposes.  

Some believe that FDA's guidance document requires specific methods. What it actually says is that you should "determine the specificity and sensitivity of the analytical method...."

  This is a far cry from requiring specific methods. A more reasonable interpretation is that you should understand the specificity of your analytical method, and take that into consideration as you utilize that method so that it is used correctly. 

This brings us to the issue of using nonspecific methods such as TOC correctly. For simplicity, I will discuss the correct use of TOC. One FDA requirement is that the TOC appropriately oxidize and measure the organic species in the target residue.

  Therefore, you will perform analytical method validation using the residue and TOC to confirm the method's applicability. Applicability indicates that the target residue is appropriately oxidized, and that it is appropriately water soluble such that it can be measured.  

Another requirement is that any detected carbon be attributed to the target residue. The carbon in a sample may be partially from the active, excipients and cleaning agent. However, we are not allowed to apportion the measured carbon among these different sources. If we use TOC, we must consider (as a worst-case assumption) that all the carbon is because of the target residue (the active, if that is the target residue). FDA also states that you "should limit background... as much as possible." Why? Because it is just good practice to decrease the background (the TOC blank) to as low and as consistent a value as possible. This is why low TOC water, ultra-clean swabs and precleaned vials are typically used for swabbing with TOC. A final requirement is determining sample stability to confirm method applicability under expected holding conditions (post sampling, before analysis). Of course, this last requirement is relevant to any analytical method. There are other TOC requirements that are common to all analytical methods, including performing sampling recovery studies.

  The bottom line is you can use TOC, but use it correctly. 

Recognition of these myths, and their lack both of scientific and written regulatory justification, can help companies avoid unnecessary work that adds little or no value to a cleaning validation programme.  

4. D. A. LeBlanc, "Why TOC is Acceptable", 

, Cleaning Validation Technologies, 24–27 (2003).  

5. FDA, Human Drug cGMP Notes, 1st Quarter 2002, PDA Letter 

www.fda.gov/cder/guidance/cGMPs/equipment.htm

 is a consultant at Cleaning Validation Technologies, San Antonio, TX, USA.			

Performing studies to mathematically 'correlate' swab and rinse sampling values does not add any value. What's more, do not expect them to mathematically correlate.