Es ist nicht klar, wie groß die Rolle von Antibiotika https://antibiotika-wiki.de/ in den Wettbewerbsbeziehungen zwischen Mikroorganismen unter natürlichen Bedingungen ist. Zelman vaxman glaubte, dass diese Rolle minimal ist, Antibiotika werden nicht anders als in reinen Kulturen auf reichen Umgebungen gebildet. Anschließend wurde jedoch festgestellt, dass bei vielen Produzenten die Aktivität der antibiotikasynthese in Gegenwart anderer Arten oder spezifischer Produkte Ihres Stoffwechsels zunimmt.
No job name
Grazing-Angle Fiber-Optic IRRAS for in Situ Cleaning Validation
Michelle L. Hamilton, Benjamin B. Perston, Peter W. Harland, and Bryce E. Williamson
Department of Chemistry, UniVersity of Canterbury, Christchurch, New ZealandRemspec Corporation, Sturbridge, Massachusetts, U.S.A.Abstract:
ever, the combination of a grazing-angle IRRAS sampling
Grazing-angle Fourier transform infrared reflection-absorp-
head with a flexible, fiber-optic cable4 provides a convenient
tion spectrometry (IRRAS) using a fiber-optic accessory has
way to collect mid-IR spectroscopic data from contaminated
been investigated as a potential in situ technique for the
surfaces in situ. Preliminary work with aluminum surfaces
detection and quantification of contamination by active phar-
has shown that quantitative data can be obtained that compare
maceutical agents on glass and metal surfaces. Two methods
favorably with the results of an industry-standard swabbing/
were used for contamination preparation: one based on smearing a known amount of sample, in solution, onto the
This contribution presents the results of investigations of
substrate and the other by spraying the substrate with an
a spectroscopic method for cleaning validation that uses
aerosol of the analyte in a volatile solvent. Chemometric
IRRAS to detect and quantify active pharmaceutical ingre-
calibrations using partial least-squares (PLS) regression are
dients (APIs) on aluminum, stainless steel, and glass surfaces. presented and evaluated for acetaminophen on aluminum and
This work uses a Remspec fiber-optic grazing-angle head,4
glass, and ibuprofen on aluminum and stainless steel. The
which permits measurements to be performed on samples
results indicate that surface loadings of 0.05 µg/cm2 is a readily
outside the spectrometer body, thereby providing the sam-
achievable limit of detection for the IRRAS technique.
pling flexibility required for in situ applications. Experimental Section Materials. The solvents used were Milli-Q water, metha- Introduction
nol (ACS or doubly distilled solvent grade), ethanol (solvent
Clean equipment is vitally important in the pharmaceutical
grade), acetone (solvent grade), and tetrahydrofuran (ACS
industry, and cleaning validation of surfaces is therefore
grade). O-Acetylsalicylic acid (aspirin), 4-acetamidophenol
critical to many processes for the manufacture of pharma-
(acetaminophen), and 4-isobutyl-R-methylphenylacetic acid
ceuticals. Ex situ analytical methods such as swabbing/HPLC
(ibuprofen) were obtained from Avocado Research Chemi-
and rinse-water analysis have been the most commonly used
cals Ltd. and Sigma Aldrich. Sodium dodecyl sulfate (SDS)
approaches, but they can be time consuming and therefore
was obtained from BDH Laboratory Supplies, dried at 60
expensive. Consequently, there has been considerable interest
°C for 2 days and then stored in a desiccator prior to use.
in new methods for rapidly checking the cleanliness of
Metal test coupons were cut from 0.75-mm-thick aluminum
equipment before its release for use.1 In this regard, direct,
and 316 stainless steel stock. Glass coupons were cut from
in situ methods potentially have very significant advantages.
3-mm-thick window glass and were roughened on one side
Fourier transform infrared reflection-absorption spec-
to prevent reflection from the back face.
troscopy (IRRAS) at a grazing angle is a well-established
Instrumentation. A schematic of the IRRAS instrument
method for analyzing thin films on solid surfaces.2,3 Com-
is given in Figure 1. The spectra were collected by using a
mercial reflectance accessories designed to fit inside the
BrukerOptics Vector 22 FTIR spectrometer with OPUS data-
spectrometer sample chamber are obviously of limited use
collection software and Quant2 chemometrics package from
in an industrial setting where the ability to take the
BrukerOptics. A prototype Remspec SpotView grazing-angle
spectrometer to the measurement region is important. How-
head was connected to the external beam port of thespectrometer by a ∼1.5-m, 3-mm diameter, 19-fiber chal-
* Corresponding author: Email: mat@remspec.com. Telephone: +1-508-
cogenide-glass optical bundle, which transmits from 5000
(1) See, for example: Amer, G. Ensuring Successful Validation: the Logical
to 900 cm-1 with the exception of a strong H-Se absorption
Steps to Efficient Cleaning Procedures. BioPharm (Duluth, Minn.) 2001,
∼2200 cm-1 (Figure 1 inset). Off-axis parabolic
(2) Umemura, J. Reflection-Absorption Spectroscopy of Thin Films on Metallic
Substrates. In Handbook of Vibrational Spectroscopy; Chalmers, J. M.,
(4) Melling, P. J.; Shelley, P. H. Spectroscopic Accessory for Examining Films
Griffiths, P. R., Eds.: Wiley: Chichester, UK, 2002; Vol. 2, pp 982-998.
and Coatings on Solid Surfaces. U.S. Patent No. 6,310,348, 2001.
(3) Kattner J.; Hoffmann, H. External Reflection Spectroscopy of Thin Films
(5) Mehta, N. K.; Goenaga-Polo, J.; Hernandez-Rivera, S. P.; Hernandez, D.;
on Dielectric Substrates. In Handbook of Vibrational Spectroscopy; Chalm-
Thomson, M. A.; Melling P. J. Development of an In Situ Spectroscopic
ers, J. M., Griffiths, P. R., Eds.: Wiley: Chichester, UK, 2002; Vol. 2, pp
Method for Cleaning Validation Using Mid-IR Fiber-Optics. BioPharm
(Duluth, Minn.) 2002, 15, 36.
10.1021/op040213z CCC: $30.25 xxxx American Chemical Society
Organic Process Research & Development A
PAGE EST: 6.3
measurements. The coupons were washed using 25-100 mLof an appropriate solvent (water for SDS, ethanol for APIs)to remove the entire amount of the target compound thathad been applied. APIs have strong UV absorbance bands(acetaminophen
19 000 L mol-1 cm-1; ibuprofen 219 18 300 L mol-1 cm-1);thus, a GBC-920 UV/visible spectrophotometer was usedcolorimetrically to measure the amount of each compoundin the API washings, and the surface loadings were back-calculated. Since SDS has no UV/visible chromophore, analternative method was required to measure the concentra-tions in the aqueous washings; one obvious choice, HPLC,was ruled out as the available instrument used UV spectro-scopic detection. Of several other techniques investigated,quantitative 1H NMR spectroscopy gave the best combinationof sensitivity and reproducibility. One hundred microlitersof dioxane (internal standard) in phosphate buffer was addedto 400 µL of rinsate to give 0.5 mL containing 100 µmolL-1 of dioxane at pH 7.5. This solution was placed in a 5-mmNMR tube with a 3-mm D2O insert. Thirty-two transientswere collected in 9 min from an Oxford Instruments AS500
Figure 1. Schematic diagram of the instrument used for the
spectrometer using 8.1-µs, 90° radio frequency pulses, with
measurement of the IRRAS of surfaces contaminated with organic compounds. (Inset)“Single beam” response of the
1.89-s acquisition times and 15-s delays between pulses. Each
system including the chalcogenide glass used in the fiber bundle.
transient comprised 30 272 data over a sweep width of 8kHz. The 4.6-ppm H2O proton singlet was suppressed by
mirrors collimate the beam onto the sample (at an incidence
using a double-pulse field-gradient spin-echo (DPFGSE)
angle of 80° to the normal) and focus the reflected light to
technique. The transients were zero filled to 128 k data prior
an integrated Infrared Associates MCT detector. The IR
to Fourier transformation, and phasing and baseline correc-
footprint produced by the grazing-angle probe is elliptical,
tions were carried out manually using Varian VNMR
with the intensity decaying from the middle toward the edges.
software, version 6.1 C. The integrals of unresolved SDS
Sample Preparation. Two methods were used to prepare
CH2 resonances at ∼1.12 ppm and the dioxane resonance at
∼3.55 ppm were determined. A calibration curve was
(1) Smear Method. A known volume (∼1 mL) of an
produced by using standardized SDS solutions, and a
accurately prepared standard solution of the target compound
standard solution was included with each set of samples to
in a quick-drying solvent was dispensed onto the surface of
ensure that the calibration remained valid.
a test coupon and then spread as evenly as possible across
Data Collection. As implemented here, IRRAS is a
the surface during drying, using the straight edge of a soft
single-beam technique requiring sequential measurement of
plastic spatula. This method is rapid and easily executed
background and sample spectra. The spectral range was 990-
without specialized equipment. The amount of applied
4000 cm-1, and the resolution was 4 cm-1 (glass substrate)
material is readily controlled and can be calculated without
or 8 cm-1 (metal substrates). Single-channel background
the need for an independent analysis. However, coverage is
spectra (R0) were obtained from clean test coupons of the
typically uneven due to beading and puddling of the solution,
appropriate substrate by averaging over 100-200 interfer-
and it is possible for some of the applied loading to be
ometer scans at 35 or 40 kHz scanning speed. Five to seven
removed during the spreading operation.
sample spectra (R) were collected from different regions of
(2) Spray Method. A Paasche double-action, internal-mix
each loaded coupon by averaging over 25-100 scans. The
airbrush (operated with a small compressor and ballast tank)
IRRAS spectrum was calculated as log(R0/R).
was used to spray the test coupons with 5-10 mL of asolution (typical concentration ∼0.5 g/L) of the target
Results and Discussion
compound in a volatile solvent. By this method, it was
IRRAS is an open-path method where the IR beam passes
possible to generate samples with more uniform coverage,
through the atmosphere before and after the sample. This
and the surface loading (amount per unit area) could be
means that fluctuations in the composition of the ambient
varied by changing the concentration of the solution or the
atmosphere that are rapid on the time scale of the experiment
number of passes of the spray across the sample surface.
can lead to the appearance of features due to atmospheric
Samples could be loaded with more than one compound
components that absorb in the mid-IR To illustrate this
either by spraying separate solutions consecutively or by
problem, the lowest trace in Figure 2 shows a spectrum
obtained from a glass coupon loaded with 0.21 µg/cm2 of
Since the precise amount of material deposited by the
acetaminophen. The fine-structured bands above ∼1400 cm-1
spray method is difficult to predetermine, an independent
arise from the bending vibration of water vapor. These and
method was used to establish the loading after the IRRAS
other features due to atmospheric species (particularly the
Organic Process Research & Development
Figure 2. IRRAS of 0.21 µg/cm2 acetaminophen on glass. Figure 3. IRRAS (after correction for atmospheric H2O) of (Lower Trace) Fine-structured spectral features above ∼1400 acetaminophen deposited at a loading of 0.93 µg/cm2 on glass cm-1 due to atmospheric H2O. (Middle and Upper Traces) (upper trace) and aluminum (lower trace) coupons using the Results of auto-subtraction of a reference H2O spectrum and spray and smear methods, respectively. application of the Bruker OPUS atmospheric compensation, respectively. Both treatments reveal the underlying API bands.
stretching modes of CO2 at ∼2350 cm-1 and H2O in theregion 4000-3400 cm-1) were nearly ubiquitous in ourexperiments, even with minimal periods between the col-lection of the background and sample spectra. The H2O bandsare especially problematic since they occur in the regionsthat contain the majority of features from the target surfacecontaminants and their intensities are relatively large at thesensitivities required to detect surface contaminations atloadings relevant to pharmaceutical requirements. Figure 4. IRRAS spectra (after correction for atmospheric
One way of addressing this problem is to apply software
H2O) of acetaminophen on glass at loadings of 0.34 and 2.08
corrections. The most straightforward method is to subtract
µg/cm2. (Inset) Enlargement of the region containing the peak
an appropriately scaled independently measured reference
used for univariate calibration and showing the limits of integration.
spectrum. Alternatively, the Bruker OPUS package includesan “atmospheric compensation” option that ameliorates
this report, and a detailed treatment can be found in refs 2
interferences due to H2O and/or CO2 by using an undocu-
mented algorithm. While not perfect, either of these methods,
High-wavenumber API bands, such as those due to O-H
carefully applied, can remove the gross effects of the
and N-H stretches, are prominent in IR absorption and
atmospheric species, as shown in upper traces of Figure 2.
diffuse-reflectance powder spectra and would normally be
The subtraction-based correction, using a H2O spectrum
regarded as candidates for quantitative analysis. However,
measured under similar conditions, was applied to the spectra
they are much weaker in the IRRAS and could be detected
(at acceptable signal-to-noise ratios) only at loadings sub-
IRRAS spectra depend on the nature of the substrate
stantially greater than those relevant to pharmaceutical
surface from which the radiation is reflected as well as those
cleaning protocols. Instead, we have examined the CdO and
of the incident medium and the surface film through which
the radiation is transmitted. Broadly, substrates can be
Expansions of the appropriate (∼1800-1000 cm-1) range
classified as metal or dielectric, with the two groups having
of the IRRAS of acetaminophen and aspirin on glass are
vastly different optical properties. Both are used in chemical
shown in Figures 4 and 5, respectively. The API features
reactors and hence are relevant to this study. An illustration
have wavenumbers similar to those for the bands in the
of the differences is presented in Figure 3 for acetaminophen
diffuse-reflectance powder spectra, but they are inverted and
at the same loading (0.93 µg/cm2) on aluminum and glass.
have quite different relative intensities. As with all spectra
The spectrum from aluminum (lower trace) has an appear-
obtained from glass coupons, the region is dominated by the
ance similar to that of an absorption spectrum and is
∼1250-cm-1 Si-O feature, whose intensity is strongly
dominated by features characteristic of acetaminophen. In
dependent on surface loading and whose profile changes from
contrast, in the spectrum from the glass coupon (upper trace)
that of an absorption-like band to a dispersive specular-
the acetaminophen features are inverted and have intensities
reflectance band as the loadings are increased.
that are much reduced in comparison with the equivalent
The conventional method of using spectroscopic data for
features from the aluminum spectrum. In addition, the
quantitative analysis is to assume a direct functional relation-
dominant feature in the glass spectrum is a strong band at
ship (such as Beer’s law) between signal intensity (peak
∼1250 cm-1 due to Si-O modes of the substrate. Discussion
heights or band area) and sample concentration. We have
of the reasons for these differences is beyond the scope of
examined this type of univariate approach for several APIs
Organic Process Research & Development
With these data, acetaminophen loadings of 0.5 µg/cm2 andaspirin loadings of 1.0 µg/cm2 can be quantified with a 95%confidence interval of ∼25%. Improvements in precisioncould be achieved through the collection of larger data sets,but the lower detection limit is largely dictated by limitationsof the water-vapor corrections rather than the instrumentation.
Despite the apparent success of the analyses presented
above, more sophisticated approaches are required in moregeneral circumstances, including those that are relevant tocleaning validation in industrial settings.
First, linear behavior in IRRAS is more the exception than
Figure 5. IRRAS spectra (after correction for atmospheric
the rule. IRRAS features of thin films on surfaces frequently
H2O) of aspirin on glass at loadings of 0.50 and 2.95 µg/cm2.
show strongly nonlinear behavior, including variations of
(Inset) Enlargement of the region containing the peak used for
relative intensities and wavelength shifts as the surface
univariate calibration, showing the limits of integration.
loading changes. In the most general sense, the complexitiesof, and interactions between, the physical phenomena thatunderpin IRRAS are such that linear behavior should not beanticipated (except approximately over restricted ranges)even with the most simplifying assumptions about the natureof the materials at the surface interfaces. At low surfaceloadings, these considerations will be compounded bymolecular effects including surface packing and consequentchanges in the orientation of the molecular dipole momentsas the loading approaches a monolayer.3
Second (and as documented above), interferences from
Figure 6. Plot of acetaminophen IRRAS band integral (1485-
the atmospheric species (especially H2O) occur in the same
1527 cm-1) versus loading as determined by UV absorption
spectral ranges as features from the target compounds. This
colorimetry. Slope: 3.1 ( 0.1; Ordinate intercept: 0.2 ( 0.1; R2 ) 0.962.
problem can sometimes be circumvented (as above) byemploying compensation procedures. However, these workbest under the conditions that prevail in a research laboratorysnamely where spectra are obtained using model samples overa short time interval and with minimal changes in all othervariables.
Finally, in industrial settings, surface contamination
generally will involve several components with overlappingIR spectra. Under these conditions, the IRRAS intensity inany given region cannot be unequivocally assigned to justone component. Figure 7. Plot of aspirin IRRAS band integral (1587-1618
As previously demonstrated,5 chemometric modeling
cm-1) versus loading as determined by UV absorption colo-
methods based on factor analysis and partial-least-squares
rimetry. Slope: 0.79 ( 0.04; Ordinate intercept: 0.21 ( 0.06;
(PLS) fitting can be used to correlate IRRAS with surface
R2 ) 0.984.
loadings of organic contaminants with the advantages that
and SDS on glass and aluminum and present here the cases
(a) they are intrinsically multivariate approaches and can
of the acetaminophen and aspirin on glass, whose spectra
therefore be used to quantify more than one contaminant
are shown in Figures 4 and 5. The intensity parameters were
simultaneously, (b) they can automatically accommodate
chosen to be the absolute integrals of single, well-distin-
nonlinear relationships between spectra and surface concen-
guished bands centered at ∼1510 cm-1 for acetaminophen
tration data, and (c) the presence of variable interfering
and at ∼1605 cm-1 for aspirin. The ranges of integration
components (including atmospheric H2O and CO2) can be
are indicated by horizontal bars in the insets to Figures 4
accounted for by segregating their effects into separate factors
and 5. To account for baseline variations between spectra, a
in the analysis. The remainder of this contribution examines
straight line was subtracted to make the endpoints of the
aspects of the application of chemometric methods to the
quantification of surface loadings of APIs on glass and metal
Figures 6 and 7 show that the univariate peak-integral
substrates using IRRAS. The samples, preparation conditions,
data are well represented by straight-line fits at loadings
and loadings are summarized in Table 1. The chemometric
between ∼0.3 and 3 µg/cm2. The results clearly establish a
modeling was performed over the spectral analysis ranges
quantitative relationship between IRRAS intensity data and
listed in Table 1 by using the OPUS Quant2 package without
levels of these contaminants on glass surfaces, thereby
atmospheric corrections or other preprocessing, except for
extending results reported previously for metal surfaces.5
the default mean-centering in Quant2.
Organic Process Research & Development
Table 1. Sample preparation data and Quant2 calibration parameters for the chemometric models of single-component surface contamination by acetaminophen and ibuprofen on glass or metal substratesa
acetaminophen acetaminophen acetaminophen ibuprofen
a API ) active pharmaceutical ingredient; RMSECV ) root-mean-square error of cross validation; RMSEP ) root-mean-square error of prediction
Figure 10. Robustness test of the Quant2 chemometric model Figure 8. Predicted versus true surface loadings of acetami- for surface loadings of acetaminophen on glass. The points nophen on aluminum (data set Ac:Al) from a leave-one-out indicate the average predicted surface loadings for data set Ac: cross validation using Quant2. The samples were prepared by glass(B) determined by using the calibration model based on the smear method, and the true loadings were calculated from set Ac:glass(A) applied to five spectra measured from each the initial solution concentrations. sample; RMSEP and R2 are calculated from the raw predic- tions. The true loadings of Ac:glass(B) were determined by UV colorimetry.
factors are similar, but the RMSECV is lower for the glassdata (RMSECV ) 0.059 µg cm-2 for Ac:glass(A) versus0.097 for Ac:Al), perhaps reflecting improved uniformity ofthe spray preparation method.
To test the robustness of such models, a second sample
set of acetaminophen on glass (Ac:glass(B), see Table 1)was prepared independently by the spray/UV method. Thechemometric model developed from Ac:glass(A) was then
Figure 9. Predicted versus true surface loadings of acetami-
used to predict the surface loadings from the Ac:glass(B)
nophen on glass (data set Ac:glass(A)) from a leave-one-out
spectra. The results are shown graphically in Figure 10; they
cross validation using Quant2. The samples were prepared by
have a correlation factor of R2 ) 0.89 and a root-mean-square
the spray method, and the true loadings were determined by
error of prediction of RMSEP ) 0.052. UV colorimetry.
These and similar results for aspirin, ibuprofen, and SDS
Figures 8 and 9 show the results of standard leave-one-
indicate that a combination of IRRAS and chemometric
out cross validations of data sets for acetaminophen on
modeling will work for a range of compounds on both metal
aluminum (Ac:Al) and glass (Ac:glass(A)) respectively
and glass surfaces. Ideally, it would be possible to build a
prepared by the smear and spray/UV methods. The spectral
single calibration for a particular compound or set of
analysis region was approximately the same as that shown
compounds on a number of different, but related, substrates.
in Figures 4 and 5, but truncated at the low wavenumber
This is likely to be a challenging goal for dielectric substrates
end for Ac:glass(A) to exclude the Si-O feature. The cross
that have distinct absorption bands in the mid-IR but should
validation results, especially the low values of root-mean-
square errors of cross validation (RMSECV < 0.1 µg/cm2),
To test this prediction, two sets of test coupons were
are encouraging and indicate that good chemometric models
prepared by the smear method using ibuprofen as the target
can be developed from these types of spectra. The correlation
compound: one with aluminum as the substrate (Ib:Al) and
Organic Process Research & Development
Figure 11. Predicted versus true surface loading of combined Figure 13. Predicted versus true surface loading of acetami- data for ibuprofen on aluminum and stainless steel (data set nophen on glass in the presence of varying amounts of sodium Ib:Al+Ib:SS(A)) from a leave-one-out cross validation using dodecyl sulfate (SDS), from a leave-one-out cross validation Quant2. The samples were prepared by the smear method, and using Quant2. The samples were prepared by the spray method. the true loadings were calculated from the initial solution True loadings for acetaminophen and SDS were determined, concentrations. respectively, by UV colorimetry and quantitative 1H NMR spectroscopy.
As a technicality, it should be noted that the aluminum
coupons used in this experiment had a dull finish, while thestainless steel ones had a mirror finish. The effect of surfacefinish on the present application of IRRAS spectra has notyet been the subject of a systematic investigation but isexpected to be a factor that must be considered when datasets are being built for calibration.
The results presented thus far pertain to single APIs on
otherwise clean surfaces. An important issue for practical
Figure 12. Robustness test of the Quant2 chemometric model
applications of the IRRAS method will be the detection of
for surface loadings of ibuprofen on aluminum and stainless
contaminants in the presence of other materials, such as
steel. The points indicate the predicted surface loadings for data set Ib:Al+Ib:SS(B) determined by using a calibration model
excipients and cleaning materials. This is a more complex
based on set Ib:Al+Ib:SS(A). The samples were prepared by
problem than those addressed above, but one that we have
the smear method, and the true loadings were calculated from
begun to investigate using model systems. the initial solution concentrations.
Figure 13 shows the results of a leave-one-out cross-
the other using 316 stainless steel (Ib:SS). Quant2 chemo-
validation of a Quant2 calibration model built from 333
metric models were built using a broad spectral range (3835-
IRRAS spectra taken from glass coupons spray coated with
995 cm-1) but excluding the 1720-2276 cm-1 window,
independently varied amounts of acetaminophen and SDS,
which contains the atmospheric CO2 band but no features
a component of many pharmaceutical cleaning agents. The
of the target API. The models for Ib:Al and Ib:SS individu-
high correlation factor (R2 ) 0.968) and the comparatively
ally were of a similar quality to those above for acetami-
low RMSECV of 0.1 µg/cm2 indicate that it will be possible
nophen, and the associated parameters are given in the Table
to develop useful calibration models that take account of
1. The combined set (Ib:Al+Ib:SS(A)) was then processed
varying amounts of additional materials in addition to the
to yield a new model with the leave-one-out cross validation
APIs that are of quantitative interest. Work on this question
shown graphically in Figure 11 and the corresponding
and the related problem of simultaneously quantifying two
parameters summarized in the right-most column of Table
or more different materials is continuing, and the results will
1. These compare very favorably with those for the individual
models, indicating that the single calibration is sufficient forboth substrates and providing encouraging evidence that it
Conclusions
might be possible to treat metal surfaces (at least) together
The work presented in this contribution has shown that
the combination of IRRAS and chemometric modeling has
A robustness test of the combined Ib:Al+Ib:SS calibration
significant promise for the in situ quantification of contami-
model was carried out by collecting six further IRRAS
nation of glass and metal surface by APIs, even in the
spectra from each of four additional coupons prepared by
presence of interference from atmospheric absorption bands.
the smear techniques; two on aluminum and two on stainless
The case for single-component APIs is firmly established,
steel substrates (data set Ib:Al+Ib:SS(B)). The average
while preliminary results for binary mixtures on glass are
loadings obtained from the model are plotted against the true
loadings in Figure 12. The overall correlation for the
For the single-component contaminants on glass and metal
predictions (R2 ) 0.887, and RMSEP ) 0.2) indicates that
surfaces, the calibration models described here have RMS
usable, robust calibrations can be built for use on more than
errors of cross validation (RMSECV) and prediction (RM-
Organic Process Research & Development
SEP) near, or below, 0.1µg/cm2. The lowest surface loadings
chemically complicated contaminant mixtures representative,
investigated were 0.01 and 0.05 µg/cm2 for ibuprofen on
for example, of those used in industrial reactor-cleaning
stainless steel. After correction for absorption by water vapor,
protocols must also be investigated. These issues are the
the 0.05 µg/cm2 spectrum exceeds the “three-times-baseline-
focus of current investigations in our laboratories and will
noise” criterion for the limit of detection (LOD), while the
be the subject of publications presented at a later date.
0.01-µg/cm2 spectrum barely passes the same test. For
Although further work is needed to determine the limits
practical purposes, loadings of 0.05 µg/cm2 can therefore
of the technique, it is becoming clear that IRRAS has the
be taken as being a readily achievable LOD for the IRRAS
sensitivity and versatility required for in situ detection and
technique applied in the manner outlined in this report.
quantification of pharmaceutical compounds and other
Further developments are likely to realize lower limits soon.
materials at surface concentrations well within the range of
To be of genuine utility in industrial applications such as
interest for pharmaceutical cleaning validation and other
pharmaceutical cleaning validation, the IRRAS method must
be shown to be more generally applicable. The range ofsubstrates must be broadened to include other metals,enamels, and certain plastics, and investigation of the
Received for review November 29, 2004.
dependence of the spectra on the surface finish. More
PAGE EST: 6.3 Organic Process Research & Development
eines schönen Morgens brechen wir mit John, dem amerikanischen Arzt, zu einer seiner Inspektionsfahrten auf. John betreut ein Trachoma-Programm der WHO hier in Guatemala. Die Indios aus den Bergdörfern gehen nie zum Arzt. Zum einen gibt es keinen Arzt in diesen Gegenden, zum zweiten käme der Besuch eines Arztes viel zu teuer, in der Tat so teuer, dass ein Indianer sich gar nicht vorstellen kann