Ruggedness of Visible-Residue Limits for Cleaning (Part II)

Visual inspection is always a component of a cleaning-validation program, and routine visual inspections for equipment cleanliness are a crucial part of day-to-day operations in a pharmaceutical facility (21 CFR 211.67 (b)-6). Visible-residue limits (VRLs) can be established for the active pharmaceutical ingredients (APIs), excipients, and detergents processed in the facility. The VRL of the API is generally the greatest concern because the API is the most potent component of the formulation. If the VRLs are lower than the health-based and adulteration-based acceptable residue limits (ARLs) as determined by the company, then the VRL is a viable approach to assess the equipment cleaning in the facility (1).

Sample preparation and viewing parameters for VRLs have been established for pilot-plant and commercial-manufacturing facilities in previous studies (1, 2). A solution or suspension of the API in methanol was applied at different concentrations to stainless-steel coupons, thus resulting in residues of uniform size. Studying viewing parameters, such as viewing distance, viewing angle, light intensity, residue composition, and observer subjectivity resulted in the optimal viewing conditions to detect visible residues. These viewing conditions in the pilot plant were set at 18 in., 30°, and > 200 lux, respectively. For commercial facilities with larger, fixed equipment, the viewing inspection parameters can be restricted. The limits for acceptable viewing conditions were determined to be ≤ 10 ft, ≥ 30°, and > 200 lux. A discussion of the applications of VRL use and the associated risks concluded that the potential for a cleaning failure was small when a well-controlled VRL program was employed (3–5).

The subjectivity of the observers and the appearance of the dried residues remained as potential limitations to a VRL program. The original VRL work used a small group of four to six observers (1, 2). For subsequent work, a larger pool of observers determined the VRLs of APIs in development at one site. In addition, a study of VRL determinations of five APIs at multiple sites helped to address the issues of observer subjectivity as well as residue sample preparation to better define the ruggedness of VRLs for cleaning (6).

The complete set of VRL data was examined for distribution and trends (6). The original VRL data set was compared with the subsequent data set and demonstrated improvements in the procedure and technique. Refining the testing range to lower concentrations improved the procedure. It resulted in more accurate VRLs and fewer in which the lowest spotted concentration was visible. The training of personnel in VRL appearance and detection both in the lab and in the field improved the technique of visual observation. The VRLs of the APIs, excipients, and formulations showed a correlation between the VRL of the API and its formulation. Appearance of the residues depends on the amount of API residue being spotted and the volume of spotting solvent, which translates to the concentration of the resulting residue. VRL residues can appear as a uniform residue at higher concentrations, or as a ring, which is typical of lower concentrations. Methanol is consistently used as the spotting solvent because its low surface tension allows it to spread to a uniform spot size and its high volatility allows it to dry quickly. Experience gained through ongoing equipment inspections and direct soiling of manufacturing equipment with test soils confirmed that potential residue on cleaned equipment would be similar in appearance to the experimental VRL residues.

Despite the extensive ruggedness data generated, the following questions about the constant parameters of the VRL studies remain to be answered:

• What is the effect of the material of construction (MOC) on VRL determination? Intuitively, the VRL would be MOC-dependent, but to what extent? Can MOC be grouped for VRL purposes?

• What is the effect of the preparation solvent? Solvents will dry differently based on volatility, and the presentation of the resulting residue could be affected. This difference could potentially form a uniform residue, as opposed to a nonuniform ring residue.

• What is the effect of drying residue under air versus nitrogen? Certain APIs degrade while air drying, but the degradation could be an artifact of the spotting solvent and not occur during normal use. This is probably more of a chemical-degradation issue rather than a physical-appearance issue.

• What are the effects of less-than-optimal viewing conditions? VRLs were established under controlled laboratory conditions. Although equipment-inspection parameters were defined (≤ 10 ft, ≥ 30°, and > 200 lux), visual inspection of large, enclosed equipment (e.g., tanks) provides a special set of challenges. The surrounding surface compositions can influence the ability to detect a VRL, and the phenomena of this influence need to be determined.

These issues will be the subject of future work and will require significant resources in time, material, and data to arrive at meaningful conclusions. The question explored in this study was the most straightforward: whether the presentation of the residues to the observers affects the VRL determination. The original study design anticipated that the lower concentration residues would not be visible. For a high percentage of samples, however, the VRL was determined to be less than the lowest residue concentration spotted on the coupons. The consistent ability of observers to see the lowest spotted residue prompted the question of whether they saw the residue because they knew there was a residue present. This was a legitimate concern when applied to clean equipment inspection, where the expectation is that there would be no residue present.

Determinations of VRLs consistently followed the same manner of presentation for the observers. The arrangement of the soiled coupons began with the ARL, which was the lower of either the health-based limit or the adulteration-based limit (4 µg/cm² ). Residue levels decreased sequentially with a solvent blank as the last sample. The observers viewed all of the sample soils as a single group. For the majority of samples (133 of 200 = 67%), the lowest spotted residue was visible for the observers. The VRL was reported as less than the lowest residue concentration. This report resulted in a refinement of the sample preparation. The updated residue preparation, as shown in Table I, targeted lower concentrations to determine the actual VRL and provide more accurate program data. With this refinement, the observers continued to visually detect the lowest residue limit in the majority of tested compounds.

Table I: Residue target concentrations.

An alternative arrangement of the residue soils was considered to determine whether presenting the soiled coupons to the observers affected VRL determinations. This alternative arrangement consisted of a randomized presentation of several compounds at or near the determined VRL, including a number of blank coupons. The observers did not know beforehand the number of compounds tested or their concentrations, the number of soiled coupons, or the number of blanks in a 25 coupon array. A randomized coupon presentation is more scientifically justified and easier to defend.

Table II: Compound visible-residue limits (VRLs)

Method

A series of four previously studied compounds with established VRLs, listed in Table II, were diluted with methanol to sequentially lower concentrations. The coupons were prepared using 100 µL of each solution resulting in four residue levels. The target concentrations were the four lowest concentrations shown in Table I: 20 µg, 10 µg, 5 µg, and 1 µg. One residue level was spotted per coupon. The spotted samples dried as elliptical residues. The resulting residue concentrations were calculated using the two elliptical radii in Table III based on the amount of residue spotted and the area of the resulting residue.

Table III: Residue concentrations.

The coupons were arranged in a random order as shown in Table IV and Figure 1. Sixteen sample coupons and four solventblank coupons were prepared. In addition, five blank coupons brought the total matrix to 25 coupons in a five-by-five array.

Table IV: Random coupon identification grid.

The viewing parameters had been established previously: viewing distance was 18 in., viewing angle was 30°, and ambient light intensity was > 200 lux (1). Eight observers participated in the study. The observers viewed the coupon array separately from one another to avoid influencing their individual choices. The results for the random determinations are presented in Table V.

Figure 1: Random coupon presentation.

After the observations were completed for the random coupon presentation, the residues were ordered sequentially, two observers were brought back to view the coupons, and their observations were recorded (see Table V). The sequential presentation was a confirmation for the previously established VRLs because the residue preparations were, in some cases, much lower than the reference VRL and because every residue preparation is slightly different based on the amount of compound weighed and the area of the resulting residue.

Table V: Observer results.

Results and discussion

The results of the VRL determinations for both the random and sequential presentations are shown in Table VI. The random coupon presentation provided comparable results to the sequential coupon presentation. For both the sequentially and the randomly presented coupons, the experimentally determined VRLs were lower than the previously determined VRLs for three of the compounds and higher for the fourth. The lower VRLs resulted from spotting lower concentrations than in the initial VRL determination. The initial determinations were deemed acceptable because they were sufficiently lower than the swab limit or ARL (2). The sample preparation used for the more recent VRL determinations resulted in lower overall VRLs and a more rugged data set (6).

Table VI: Visible-residue limit (VRL) determination.

Aside from the presentation of the coupons, the other significant difference was the number of blank coupons presented. The VRL determination using the random coupon presentation involved a greater number of blank coupons (i.e., nine) than did the sequential coupon presentation (i.e., four). It was felt that the additional blank coupons would increase the physical separation of the soiled coupons, limit direct side-by-side comparisons of the soiled coupons, and enable a more independent assessment of each of the individual soiled coupons.

Figure 2: Sequential coupon presentation.

The consistent results obtained using either method of coupon presentation answered a long-standing question. The sequential coupon presentation raised a concern that the decreasing residue concentrations and the close proximity of adjacent residues might lead observers to see residues where they might not ordinarily have identified a visible residue.

Because the two methods of coupon presentation result in the same VRL, which way should one present coupons for VRL determinations? The sequential coupon presentation offers marginal advantages in the following areas.

• It provides ease of setup and recording. The random coupon presentation requires slightly more paperwork with its matrix approach.

• It is easier to do one or more multiple compounds/residues at a time. The random approach is more suited to multiple determinations.

• It makes more efficient use of coupons, because the random approach will use additional coupons to fill in the square matrix as necessary.

As noted above, the primary advantage of the random presentation was to increase the physical separation of the soiled coupons and limit direct side-by-side comparisons.

To ensure rugged, reproducible results, however, the observers must be thoroughly trained. It is always emphasized during training that the VRL is a physical property of the residue and that the ability to see every spotted residue is not an expectation. Observers are then more likely to report what they see rather than what they think they are expected to see. Therefore, the optimal coupon presentation depends on the level of training for the coupon preparation and the observers. For newly trained personnel, the random coupon presentation might be appropriate, and the sequential coupon presentation might be more appropriate for more experienced personnel.

Conclusion

The presentation order of the coupons appears to have no effect on the determination of VRLs. Both a sequential presentation from high to low concentration and a random presentation resulted in the same conclusions for VRL levels, thus confirming the ruggedness of the VRL determination.

Richard Forsyth is a senior consultant with Hyde Engineering + Consulting, 6260 Lookout Rd., Suite 120, Boulder, CO 80301, tel. 303.530.4526, fax 303.581.0839, richard.forsyth@hyde-ec.com.

Submitted: Oct. 1, 2010. Accepted: Dec. 1, 2010.

References

1. R.J. Forsyth and V. Van Nostrand, Pharm. Technol. 28 (10), 58–72 (2004).

2. R.J. Forsyth and V. Van Nostrand, Pharm. Technol. 29 (10), 152–161 (2005).

3. R.J. Forsyth and V. Van Nostrand, Pharm. Technol. 29 (4), 134–140 (2005).

4. R.J. Forsyth, J.L. Hartman, and V. Van Nostrand, Pharm. Technol. 31 (4), 134–140 (2007).

5. R.J. Forsyth et al., Pharm. Technol. 30 (11), 90–100 (2006).

6. R.J. Forsyth, Pharm. Technol. 33 (3), 102–111 (2009).