API versus excipient versus formulation determinations.
VRL data were broken into API, excipient, and formulation categories and analyzed to determine VRL correlation between a formulation
and its components. Earlier work (1) compared VRL of 12 formulations with VRLs of the formulation components. Logically, the
VRL of the formulation would be the same as the lowest component VRL—this was the case in 7 of the 12 comparisons. In three
of the cases, however, the formulation VRL was higher, and in the other two cases, it was lower than the component VRLs.
The more important comparison was between VRL of the formulation and its API. In a pilot plant, it is not practical to perform
VRL on every development formulation because the formulation compositions continually evolve up to the final market formulation
selection. In the VRL comparison of formulations against components, 9 of the 12 formulation VRLs were lower than the VRL
of the respective API. In one case, they were equal, and in the remaining two the formulations, the VRL was higher. The data
concluded that VRL of API is not a good indicator for VRL of a formulation.
The data, however, were generated during the original VRL work where the residue concentrations and observer variability were
higher. The subsequent VRL work generated additional data with increased experience and refined technique. The data gap narrowed
between the formulations and APIs. The final average VRL of the 64 formulations was 0.7 μg/cm2. Of the 113 API determinations, it was 1.0 μg/cm2. Average VRL of the 64 excipients tested to date was 1.6 μg/cm2, and data showed significant overlap among formulations, APIs, and excipients.
A t-test comparison of API and formulation VRL data in Figure 2 shows that data distributions were equivalent. The formulation
VRL data, despite its lower average value, was not statistically different when compared with API VRL data. The expanded data
set analysis demonstrates that VRL of API is a good indicator for VRL of a formulation. VRL of a development API can be determined
and safely used as the VRL for the development formulations.
Figure 2: Visible-residue limit (VRL) distribution.
VRL data from the three sites are shown in Table IV. Data from the Hoddesdon facility was generally lower than data from the
other sites. Data from the West Point facility was slightly higher, which correlated to smaller spot sizes and resulted in
higher spot concentrations (see Table V). Observers in Hoddesdon and West Point typically detected the lowest or next-to-lowest
residue level. Data from the Montréal site resulted in three VRLs that were higher than VRLs from other sites; observer variability
at Montréal was also greater. Of the three higher levels, one was comparable with the established VRL, and the other two were
higher. All three of the higher levels were still well below the adulteration limit of 4 μg/cm2. A review of observer data showed that, in all cases, the higher levels were based on one observer not detecting the residue.
Otherwise, the data more closely agreed with that of the other sites.
Table IV: Multisite visible-residue limit (VRL) data.
Several factors led to variability in the multisite data. The sample solution concentrations, spot sizes, and the resulting
residue concentrations influenced VRL determination. The Hoddesdon site's lowest residue level was lower than the other sites,
explaining their overall lower VRL levels. Observer variability at the Montréal site was similar to the early West Point data;
a single observer skewed the results compared with the other observers and sites.
Table V: Multisite residue concentration comparison.
Overall, VRL determination was comparable at all three sites, and the experimental variability from sample preparation and
observer subjectivity posed no risk for a potential cleaning failure because all VRL values were well below the ARL. The study
also highlighted the value of the VRL training program and the experience gained through ongoing visual equipment inspections.