A risk-management assessment of VRL applications includes the identification of potential risks. The potential risk analysis
determined the probability of occurrence and seriousness. Probabilities occur in low, medium, and high categories. Likewise,
risk seriousness has designations of low, medium, and high. Individual risks populate a matrix of probability versus seriousness.
Risk evaluation leads to risk management. Risk avoidance takes the necessary steps to prevent a risk from occurring. Risk mitigation lessens the probability and seriousness of the risk. Risk acceptance is appropriate if the probability of occurrence is low or will not be serious enough to compromise product quality.
The most serious risk using VRLs is the potential that dirty equipment passes visual inspection and the subsequently manufactured
formulation is compromised. Another risk is a regulatory agency challenge to the VRL approach. Finally, the subjectivity of
visual assessment is a risk. The closer a VRL is to the ARL, the greater this risk becomes.
Risk analysis: key terminology
In addition to these risks, there are limitations when applying VRLs. To date, VRL determinations have been limited to stainless
steel surfaces, which comprise the vast majority of equipment surfaces. Other materials of construction were not evaluated
because of their poor reflective properties and would have to be addressed separately. VRL applications also have obvious
limitations with respect to assessing microbiological control, particularly with wet-processing equipment. A cleaning-validation
program would still need to assess the risks for microbial contamination of equipment. Surface sampling and rinse testing
would be required to demonstrate a satisfactory state of control with respect to bioburden.
Several causes are possible for dirty equipment passing visual inspection and compromising the subsequently manufactured formulation.
The most likely scenario is that the inspector either did not perform a 100% inspection or performed an inadequate inspection
on the equipment. As shown in Figure 1, the seriousness of an inadequate inspection is high because the subsequent batch could
be compromised. The range of probability of an inadequate inspection depends on several factors. Training of the inspectors
is a crucial component if VRLs are used to determine equipment cleanliness. Each active pharmaceutical ingredient (API) has
a characteristic VRL, and inspectors must be familiar with the appearance of each residue. Visual inspection must be performed
appropriately by inspecting all contact surfaces of the product with proper viewing angles of lighting and equipment. Documented
training must be updated as new APIs and formulations are introduced into the manufacturing matrix.
VRLs resulting in residues above the ARL are a potential risk. A comparison of the VRL relative to the ARL is essential before
any use of VRLs is initiated. The VRL must be lower than the ARL to use the VRL for cleaning evaluation. Although the seriousness
for this risk is high, the probability is low.
The risk of a regulatory agency challenge is more likely when using VRLs only if there is not an initial validation study
incorporating surface or rinsate testing. FDA has stated that the use of visual examination is limited between batches of
the same product (2). Agencies from the European Union, Japan, and Canada also are likely to challenge a cleaning validation
program using VRLs. The probability of an agency questioning the use of VRLs is high. The seriousness of the risk depends
on the extent of VRL use and the data generated to support VRL use. VRL implementation can range from very specific situations,
to more general applications, to the basis for an entire cleaning validation program. The data to support the use of VRL can
range from a single, general limit (7, 8) to specific VRL determinations for each API and drug product (4, 6).
The subjectivity of VRL determinations and routine visual inspections are an ongoing risk. The probability and seriousness
of this risk depend on the extent of training, ongoing monitoring, and VRL data generation, which can be minimized in a well-run
VRL program by using at least four observers to determine VRLs and redundant monitoring, resulting in low probability and