PEER-R EV IEW ED

Cleaning Validation: Factors

Affecting Recovery Data and
Material of Construction
Grouping
Richard J. Forsyth

ABSTRACT The residue assay methods are typically validated for

Material of construction is a factor in recovery of resi- the following parameters:
due for cleaning validation. Analysis of existing recov- • Linearity
ery data demonstrated that recovery factors for drug • Precision
products on different materials of construction could • Sensitivity
be categorized into several groupings. The groupings • Specificity
based on the recovery data were not aligned with the • Recovery.
material composition (e.g., metal, plastic, glass, etc.). An
examination of additional factors clarified the grouping From an analytical standpoint, recovery is from the
of materials versus recovery data. cleaning test sample, usually from a swab. For the clean-
Materials of construction used in laboratory studies to ing program, the concern is the recovery of the residue
represent commercial manufacturing equipment must be from the manufacturing equipment (1, 2). Recoveries
representative of equipment. Materials of construction are determined through experiments in which sample
that exhibit low recovery should be replaced wherever equipment materials spiked with known amounts of the
practical for an alternative material with a higher recov- substance of interest are swabbed and tested. The swab
ery to avoid any potential cross-contamination due to and the swabbing solvent must be capable of recovering
slow release of residue from the material of construction. a sufficient amount of material to allow an accurate and
If problematic materials cannot be released, equipment precise measurement of the spiked component.
should be dedicated or restricted for use to the active The most important aspects for product recovery
pharmaceutical ingredient with low recovery. factors are that the data are consistent, reproducible,
and provide an adjusted ARL that is above the limit of
INTRODUCTION quantitation of the analytical method. The ARL must
Residue assays are a critical requirement in establish- be achievable and practical. If recoveries are too low,
ing a validated cleaning program. They are essential to either the methods need to be optimized or the manu-
accurately determine amounts of residual active phar- facturing equipment must be dedicated or restricted for
maceutical ingredient (API) remaining on equipment manufacturing only the specific API.
after cleaning. This determination is then compared to A recent study (3) gathered and statistically analyzed all
the acceptable residue limit (ARL) for a given process or available historical data and achieved the following:
equipment train (1). • Grouped materials of construction according to recov-

For more Author

information,
go to
gxpandjvt.com/bios
[
ABOUT THE AUTHOR
Richard Forsyth is a pharmaceutical consultant with a background in cleaning validation and analytical
chemistry. He may be reached by e-mail at [email protected].

gxpandjv t.com Journal of Validation T echnology [Autumn 2009] 91

PEER-R EV IEW ED

ery performance. This resulted in five distinct group- adequately controlled to minimize subjectivity and
ings with statistically different recovery levels. These should consistently recover enough residue so that a pre-
groupings, however, could not be correlated to mate- cise measurement is assured. The combination of swab
rial properties. For example, different types of plastics material and recovery solvent should not interfere with
were noted in all five groups. Most metals, including the subsequent sample assay.
stainless steel, were categorized into the two groups Although the composition of the materials of construc-
with the highest recovery levels; however, aluminum tion provided no correlation with the recovery data, the
was in the second lowest recovery group. physical properties of the materials could affect recover-
• Selected representative materials from each grouping ies. There have been studies that have examined surface
for potential testing if appropriate. roughness and how it affects the ability to clean surfaces
(6). The hardness of the individual materials might also
The end result of the analysis was a reduction in the help explain certain residue recoveries. The porosity of
number of studies required for new substances in a clean- certain materials could account for several of the low
ing validation program. The analysis concluded that a recovery data sets.
recovery study conducted at one site using stainless steel The objective of this study was to gather historical data,
would serve as a representative material of construction generate necessary additional data, and statistically ana-
for most materials used in drug product manufacturing, lyze the data relating to these parameters. The end result
and is applicable across multiple sites. of this analysis identified the physical characteristics of the
The study was not able to identify the physical character- materials of construction and parameters that influence
istics of the materials and parameters that influence recov- recovery data results. An understanding of the relationship
ery data results. There was no obvious correlation between between the recovery data and the materials of construc-
the recovery data and the similarity in composition (e.g., tion was achieved.
metals, plastics, glass) of the materials of construction.
Different metals, rubbers, and plastics occupied different EXPERIMENTAL
groups across the range of residue recoveries. Because the The solubility of the recovered materials was considered
data set was generated at numerous sites worldwide over as a factor affecting swab recovery. The solubility of a
several years, it was not practical to assemble representative number of APIs was tabulated and analyzed in compari-
samples of all of the materials of construction. son to the respective recovery data. The swab technique
The following are a number of additional parameters was also examined as a factor in recoveries. The swab
that could affect the recovery of residue from equipment techniques from the different sites were reviewed and the
surfaces (4): swab materials were examined for possible correlation
• Residue solubility to low recoveries. The recovery solvent and extraction
• Swab material procedure were examined for several low recovery APIs
• Solvent type to determine if recoveries could be raised.
• Recovery technique by sampling technician The cleanability data for a number of APIs and formula-
• Residue cleanability (5), which in turn may be related tions were tabulated, statistically analyzed, and compared
to its solubility. to the recovery data. The gravimetric-based cleanability
method consisted of placing a wet slurry consisting of a
These parameters are interrelated. The cleanability of set amount of water and formulation and API onto a pre-
the residue impacts the choice of cleaning procedure; the weighed coupon (23 mm x 20 mm). The coupon was 316L
harder a residue is to clean, the higher the anticipated SS foil. In a single run, five soiled coupons were made
amount of residue to be swabbed. The amount of antici- along with three control coupons, and were placed into a
pated residue affects the choice of swab and solvent as well humidity-controlled box (RH = 30%) for a set amount of
as the validation of the analytical testing method. time (dirty hold time, usually 24 hrs). The coupons were
The materials used to recover the residue also affect then weighed and photographed and then placed into a
recovery. The swab material must be able to absorb suf- modified cuvette holder.
ficient residue and solvent to remove the residue from The coupons were then exposed to a cleaning cycle that
the equipment material surface. The type and amount usually consisted of three process steps including pre-rinse,
of recovery solvent must dissolve the residue sufficiently detergent wash, and final rinse. The typical cycle consisted
for removal without leaving residue or solvent behind. of a 10-second dip in 80°C USP water; a 10-second dip in
The swabbing technique should be standardized and 80°C clean-in-place (CIP) solution (0-3%), and a last 10-
92 Journal of Validation T echnology [Autumn 2009] iv thome.com
 R ICH A R D J. FORSY T H

seceond dip in 80°C USP water. The coupons were then Table I: Solubility and recovery data comparison.
returned to the glove box for an additional 24 hours. After Compound Recovery Water solubility
which, the coupons were weighed and photographed. The Clinoril 73 1000
cleanability was then calculated based on weight.
Invanz 98 500
Recovery experiments were conducted with represen-
tative materials of construction that had provided both Singulair 90 240
high and low recoveries. Earlier recovery data (3) defined Crixivan 76 100
groups of material of construction inclusive of variability Tryptanol 75 100
due to individual compounds. Therefore, the groupings Prinivil 99 97
were independent of the compound recovered. For this Fosamax 87 40
study, two compounds were tested to provide adequate
Fosamax Plus D3 75 40
data points for each material of construction. Stainless
steel and a smooth Neoprene rubber coupons not only Cpd D 82 25
represented the high and low recoveries (3), they also Niacin 87 4.25
represented a hard and soft material, respectively. Stain- Sinemet 81 1
less steel coupons machined to different roughness factors Pepcid 77 0.74
provided data. A Surtonic Duo surface finish analysis
Compound C 83 0.722
instrument (Taylor-Hobson, ID C-5524) was used to mea-
Compound A 84 0.19
sure surface roughness of the coupons. Coupons with
relatively smooth surfaces presented a uniform roughness. Moduretic 5/50 mg 87 0.1
Rough surface coupons had grains resulting in different Proscar/Propecia 91 0.1
roughness measurements in perpendicular directions. Compound B 98 0.1
In these cases, unidirectional swab samples attempted to Decadron 89 0.1
discern recovery differences between the two roughness
Noroxin 86 0.091
measurements. A rough Neoprene rubber and anodized
Arcoxia 83 0.06
aluminum coupons provided data to address porosity.
Residue spots were prepared in triplicate and individual, Compound E 91 0.03
single swabs taken for recovery. The swab samples were Zocor 81 0.03
assayed with a validated high performance liquid chro- Stocrin 78 0.01
matography (HPLC) method specific for the API and an Mevacor 40 mg 81 0.004
average recovery factor was determined.

RESULTS AND DISCUSSION

Results are discussed for API aqueous solubility and recov- adequate solubility in the chosen recovery solvent. The
ery; recovery, aqueous solubility, and cleanability; material relatively high recovery data demonstrated that acceptable
hardness; surface roughness; and porosity. laboratory methods using appropriate recovery solvents
had been developed. Depending on recovery data, recov-
API Aqueous Solubility and Recovery ery solvents may use combinations of polar, semipolar,
The initial analysis compared the previously determined and non-polar solvents.
recovery data (3) with the solubility data for the respective The recovery factors including swab technique, swab
compounds (see Table I). Solubilities ranged from 1000 material, recovery solvent, and extraction procedure were
mg/mL to 0.004 mg/mL. Recoveries for the compounds examined as part of the earlier study (3). These four factors
examined ranged from 89% to 73%. There was no direct were combined and considered as the site-to-site vari-
relationship between the water solubility of the compound ability for the recovery data. The additional variability
and its respective recovery. In fact, the compound with among sites was not significant relative to the variability
the lowest recovery (73%) also had the highest solubility for repeated measurements due to differences in product
(1000 mg/ml). This was not unexpected for this data set. and material of construction. Ninety-seven percent of
Although water is the first choice as a recovery solvent, if the variability was attributable to different recovery fac-
water solubility is low an alternate organic solvent (e.g., tor (RF) means for each material and product, while only
ethanol, methanol, acetonitrile) is employed. Part of 3% additional variance was attributable to the swab, sol-
the recovery assessment is to assure the compound has vent, and technique factors across sites. This meant that
gxpandjv t.com Journal of Validation T echnology [Autumn 2009] 93
PEER-R EV IEW ED

Table II: Cleanability and recovery data comparison.

Cleanability
Compound Recovery Water solubility Water CIP 100 CIP 300
Fosamax Plus D3 75 40 100 100
Mevacor 40 mg 81 0.004 99 100
Stocrin 78 <0.010 99 100
Compound E 91 <0.03 73 90
Arcoxia 83 0.06 68 96 78
Pepcid 77 0.74 61 69 91
Compound A 84 0.19 44 45 50
Zocor 81 <0.03 40 42 37

Table III: Recovery, solubility, and cleanability aqueous and differed only in proprietary formulation.
comparison in descending order. CIP 100 is an alkaline cleaning liquid; CIP 300 is a neutral
phosphate-free cleaning liquid.
Recovery Water solubility Cleanability
Table III more clearly demonstrates the lack of corre-
Compound E Fosamax Plus D3 Fosamax Plus D3 lation among the data. Recovery, water solubility, and
Compound A Pepcid Mevacor 40 mg cleanability are listed in descending rank order for the eight
Arcoxia Compound A Stocrin compounds with data in all three categories. There is no
Mevacor 40 mg Arcoxia Compound E correlation among the data and there is no obvious relation
between the solubility and cleanability data, which might
Zocor Compound E Arcoxia
be reasonably expected. The only conclusion is that there
Stocrin Zocor Pepcid
is no discernable connection or trend among the data.
Pepcid Stocrin Compound A An analysis of the HPLC recovery data led to a number
Fosamax Plus D3 Mevacor 40 mg Zocor of conclusions. An initial comparison in Table IV of the
recovery data to the previously determined recovery data
(3) demonstrated similar recovery data for all materials
across-site and within-site variability could be combined, except the Neoprene samples. The Neoprene samples
and comparisons of materials and products across sites for this study had been sourced from a different vendor
had essentially the same precision as comparisons made than the samples for the original (3) work. The Neoprene
within a single site. This was a very useful finding, since in the original work was not designated as smooth or
it allowed RF results to be leveraged among different sites rough. This demonstrates the importance of performing
without accounting for site-to-site differences. recoveries on the same material as used in manufacturing
whenever possible.
Recovery, Aqueous Solubility, and
Cleanability Material Hardness
A similar analysis compared the recovery data with Comparison of the percentage of recovery data for stain-
cleanability data for several of the compounds generated less steel and smooth Neoprene in Table IV indicates that
in both water and detergents in an independent study the hardness or softness of the material had no impact on
(see Table II). For this data set, the cleanability ranged recovery. The average recoveries from the mill finish stain-
from 100% to 37%. The respective recovery data ranged less steel and the smooth neoprene were equivalent.
between 91% and 75%. Again there was no direct rela-
tionship drawn from the data. Although cleanability is Surface Roughness
a good indicator of the ability to clean a compound from Table V provides recovery data from different surface rough-
manufacturing equipment, it does not necessarily indicate ness materials. Data from stainless steel were equivalent
the ability to recover residue for testing. Table II data dem- across the entire roughness range tested. The smoothest
onstrates general consistency in rank order cleanability coupons (0.01 µm) were mirror finish stainless and the
for the three cleaning liquids (water, CIP 100, CIP 300). 0.83 µm were mill finish or 316 finish stainless steel. The
This may be expected because all cleaning liquids were rougher stainless steel finishes (1.1 – 4.5 µm) were specifi-
94 Journal of Validation T echnology [Autumn 2009] iv thome.com
 R ICH A R D J. FORSY T H

cally machined for these experiments. The glass, brass, Table IV: Material % recovery.
polytetrafluoroethylene (PTFE), and smooth Neoprene, Material Average % recovery Average % recovery
which were comparatively smooth, had recovery num- reference 3
bers in the same range as the stainless steel. The rubber Stainless steel 81.8 83.8
with a 4.94 µm roughness also demonstrated equivalent Glass 81.1 85.5
recovery data. Recoveries from the rough Neoprene were
Brass 80.9 80.6
much lower, indicating that surface roughness is a factor
in the ability to recover residue from equipment surfaces. PTFE 85.5 83.9
However, recoveries from the aluminum sample were Neoprene (smooth) 79.4 31-44
also low even though the roughness was comparable to Aluminum #7 55.3 53-56
stainless steel and rubber. This indicated that another Rubber #5 88.6 81.9
factor was the cause of the low recoveries.
Neoprene (rough) 11.7 31-44

Porosity
The rough Neoprene was also very porous and this physical Table V: HPLC recovery data.
characteristic certainly contributed to the low recoveries. Surface Amount
The spotted residue clearly entered the pores of the mate- roughness spotted Average %
rial as soon as it was spotted. A background check of the Material (µm) (µg) recovery
aluminum sample revealed that it was anodized alumi- Stainless steel 0.01 100 86.6
num, which is also a porous material (7, 8). Anodized Stainless steel 0.4 100 77.6
aluminum can have pore diameters from 10–500 nm and
Stainless steel 0.83 100 83.4
with different layer thicknesses.
Other materials demonstrating lowest recoveries were Stainless steel 1.1 100 83.7
plastics containing methacrylate or butadiene-acrylonitrile Stainless steel 1.3 100 79.2
(3). RF for these materials ranged from 31-44. Understand- Stainless steel 3.7 100 84.1
ing the recovery factors for cleaning validation of elastomers Stainless steel 4.5 100 78.3
(rubbers) and plastics, both uncrosslinked and crosslinked,
Glass 0.03 100 81.1
may be complex. Material contamination during the
Brass 0.16 100 80.9
manufacturing process and during the cleaning process
are affected by the transport properties for the pairing of PTFE 1.11 100 85.5
the particular compound and material of construction Neoprene 1.57 100 79.4
being considered, including the time of contact between (smooth)
the compound and material of construction, and the time Aluminum #7 2.3 100 56.4
of recovery used for the cleaning validation. The transport Rubber #5 4.94 100 88.6
properties include permeability, the diffusion coefficient, Aluminum #7 5.34 100 54.1
and the solubility coefficient of the compound into the
Neoprene (rough) 12.8 100 13.5
material of construction. The permeability is essentially
the product of the diffusion coefficient and the solubility Neoprene (rough) 18.9 100 9.9
coefficient, and these three properties are temperature
dependent. A higher diffusion coefficient indicates that Transport of compounds through materials generally
a compound will diffuse further into the material of con- occurs more rapidly in elastomers such as Neoprene,
struction, and a higher solubility coefficient means that EDPM, latex, or silicone. Material transport is relatively
more compound can equilibrate within the material of less in glassy thermoplastics such as Lexan. The degree
construction at a given concentration of exposure. Another of crystallinity, the degree of crosslinking, and the pro-
factor that can influence the recovery is the solvent chosen portion of fillers will also affect the transport properties.
for the cleaning validation, as the solvent can affect the The longer period of time a material is exposed to a com-
transport properties of the material of construction. If the pound, the further into the material the compound will
material of construction has a high affinity for the solvent, diffuse, thereby making full recovery of the compound
the material could swell and the compound under test more difficult for a fixed cleaning period of time. Glassy
could further diffuse into the material along with some of polymers with glass transition temperatures above room
the cleaning solvent, thereby reducing the recovery. temperature tend to allow materials to diffuse more slowly
gxpandjv t.com Journal of Validation T echnology [Autumn 2009] 95
PEER-R EV IEW ED

than rubbery polymers with glass transition temperatures any potential cross-contamination due to slow release of
below room temperature. However, rubbery polymers are residue from the material of construction. If replacement is
usually crosslinked and filled, which tends to decrease not possible, equipment should be dedicated or restricted
diffusion rates within the given rubbery material. for use to the API that had low recovery.
Overall, transport properties are very specific. Com-
plex interactions dependent upon the solvent, material REFERENCES
of construction, and the compound under test cannot 1. FDA, Guide to Inspection of Validation of Cleaning Processes,
be generalized in a manner that will allow prediction of Division of Field Investigations, Office of Regional Opera-
recovery during cleaning validation tests. Explanation tions, Office of Regulatory Affairs, Washington, D.C. July
of the relative performance of one compound against 1993.
another for a given material of construction or the rela- 2. R. J. Forsyth and D. Haynes, “Cleaning Validation in a Phar-
tive performance of one material of construction against maceutical Research Facility,” Pharm. Technol. 22 (9), 104
another for a given compound cannot be predicted. In – 112, 1998.
order to better understand particular outcomes, specific 3. R. J. Forsyth, J. C. O’Neill, and Jeffrey L. Hartman, “Cleaning
measurements of fundamental transport properties for Validation: Grouping Materials of Construction Based on
the solvents, compounds, and materials of construction Recovery Data,” Pharm. Technol., 31 (9), 104 – 112, 2007).
would be necessary. 4. G. M. Chudzik, “General Guide to Recovery Studies Using
These measurements are unnecessary for the established Swab Sampling Methods for Cleaning Validation,” J. Val.
materials of construction for which average recoveries Technol., 5 (1), 77 – 81, 1999.
have been determined, regardless of the compound recov- 5. R. Sharnez, J. Lathia, et al., “In Situ Monitoring of Soil Dis-
ered and the solvent employed (3). If a new material of solution Dynamics: A Rapid and Simple Method for De-
construction is introduced, an alternative to the specific termining Worst-case Soils for Cleaning Validation,” PDA J
measurements above is to conduct recovery studies with Pharm. Sci. and Technol, 58 (4), Jul-Aug 2004.
one to three different API and experimentally determine 6. F. Riedewald, “Bacterial Adhesion to Surfaces: the Influence
into which recovery group the new material of construc- of Surface Roughness,” PDA J Pharm. Sci. and Technol, 60 (3)
tion falls. 164 – 171, May-June 2006.
7. V. Sokol, I. Vrublevsky, et al., “Investigation of Mechani-
CONCLUSIONS cal Properties of anodized Aluminum using Dilatometric
Material of construction is a factor in recovery of residue Measurements,” Anal. Bioanal. Chem. 375 968 – 973, March
for cleaning validation. More specifically, the ability of 2003.
the residue to permeate into the surface of the material 8. E. S. Koolj, H. Wormeester, et al., “Optical Anistropy and
lowers recovery. Although the solubility of the residue Porosity of Anodic Aluminum Oxide Characterized by
in the recovery solvent, the ability of the swab to recover Spectroscopic Ellipsometry,” Electrochem.Solid-State Lett. 6
the residue, and the sampling technician swab technique (11) B52 – B54, 2003. JVT
also contribute to recovery data, these factors have been
standardized to demonstrate recoveries of >75% for many ARTICLE ACRONYM LISTING
materials of construction. API Active Pharmaceutical Ingredient
The use of new materials of construction for pharmaceu- ARL Acceptable Residue Limit
tical manufacturing should be evaluated on an individual CIP Clean-in-Place
basis to determine the ability to recover pharmaceutical HPLC High Performance Liquid Chromatography
residue from its surface. Materials of construction coupons RF Recovery Factor
used in laboratory studies to determine recovery data must USP United States Pharmacopeia
be exact replicas of materials used to fabricate equipment.
The data of this paper demonstrated that all Neoprene was
not the same—there were significant recovery differences
between “smooth” and “rough” Neoprene. High surface
roughness materials as exemplified by rough Neoprene
had the lowest % recovery data. Materials of construction
that exhibit low recovery should be replaced wherever
practical. Replacement materials should be evaluated to
ensure a higher recovery of the API of interest to avoid
96 Journal of Validation T echnology [Autumn 2009] iv thome.com