Rapid Headspace Oxygen Analysis

for Pharmaceutical Packaging Applications

Allen C. Templeton,* Yieng-Hau R. Han, Rajiv Mahajan, Rey T. Chern, and Robert A. Reed

O
xidative degradation of drug substances in pharma-
ceutical formulations is well documented (1). Although
exact mechanistic details about what promotes reac-
tions between drug substances (RH in the schematic
The accurate below) and molecular oxygen in pharmaceutical formulations
determination is not understood fully, such reactions are generally thought to
of headspace be in the category of auto-oxidation processes as described by
oxygen levels the following schematic (2):
in common Initiation RH R

STEVE SMITH

pharmaceutical Propagation R

O2 ROO

packaging
ROO

RH ROOH
R

configurations
presents several analytical challenges in Termination 2ROO
molecular products
terms of both sampling and measurement.
This article reviews headspace gas Molecular oxygen is involved in the propagation step of this re-
action, and it is an integral part of the catalytic cycle that gen-
chromatography, electrochemical analysis,
erates the oxidative degradation of drug substances in pharma-
frequency-modulation spectroscopy, ceutical formulations.
fluorescence quenching, and quadrupole Oxidative degradation of the active drug substance in a for-
mass spectrometry as methods available to mulation decreases drug potency and reduces product shelf
packaging R&D technologists and life (2). The rate of oxidative degradate formation is propor-
laboratory analysts for rapid headspace tional to the structural susceptibility of the specific drug sub-
stance to auto-oxidize and to storage (reaction) conditions
oxygen analysis. The article describes the
such as temperature, humidity, oxygen concentration, and time.
strengths and limitations of each approach. Other deleterious effects of oxidative processes include prod-
uct discoloration, changes in dissolution rate and profile, pre-
cipitation, and the generation of foul odors and flavors (1).
Allen C. Templeton, PhD, is a senior Most important, oxidative degradation products generated in
research chemist in the pharmaceutical the final pharmaceutical product during storage may also have
analysis and control department at Merck adverse pharmacological properties, including those associ-
Research Laboratories, WP78-210, ated with toxicity or negative side effects.
Sumneytown Pike, West Point, PA 19486,
Drug oxidation can also pose significant challenges during
tel. 215.652.9204, fax 215.652.2835,
[email protected]. Yieng-Hau manufacturing processes, particularly for easily oxidized drug
R. Han is a research associate in the substances. The effects are especially pronounced in formula-
pharmaceutical analysis and control tions that require manufacturing unit operations to be per-
department; Rajiv Mahajan, PhD, is a formed in solutions or suspensions. In such cases, extra care
senior research chemical engineer and Rey
may be required to minimize or eliminate exposure to oxygen.
T. Chern, PhD, is a senior research fellow
in the packaging systems development Inert-gas purging to lower solution oxygen levels as well as inert
department; and Robert A. Reed, PhD, blanketing of process vessels have long been used in parenteral
is a director in the pharmaceutical analysis manufacturing processes. Besides purging at key manufactur-
and control department, all at Merck ing steps — for example, during drug compounding or while
Research Laboratories.
holding a drug in solution before lyophilization — packaging
*To whom all correspondence should be addressed. with an inert blanket over the final product also protects paren-
teral formulations that are susceptible to oxidation. Removing
44 Pharmaceutical Technology JULY 2002 www.phar mtech.com
one of the two key reactants in the auto-oxidation cycle from are beyond the capabilities of a single measurement technique.
the package may lengthen product shelf life and ensure prod- This article describes various measurement options that are
uct quality when it reaches the consumer. Modified atmosphere available to packaging R&D technologists and laboratory an-
packaging (MAP), a process used to package oxygen-sensitive alysts in the context of pharmaceutical packaging applications.
commercial products such as foodstuffs and medical devices in The article is organized by technique, with a brief introduc-
an inert atmosphere, can be used to increase shelf life. Although tion to measurement principles followed by a discussion of
MAP has taken hold in the parenteral pharmaceuticals indus- pharmaceutical packaging applications.
try, relatively few solid dosage forms are packaged under re-
duced oxygen levels. Experimental
Smart packaging methods such as MAP may advance the com- Sample preparation. Standards that contained various concen-
mercial development of oral formulations with drug compounds trations of oxygen in nitrogen were prepared from double gravi-
that are particularly prone to oxidation and thus make available metrically certified gas cylinders (Scott Specialty Gases, Plum-
critical therapeutic agents in a high-quality, stable, and conve- steadville, PA). The concentrations were 0.00, 5.0, 10.0, 15.0,
nient dosage form. Integrating formulation and packaging im- and 20.0% oxygen in nitrogen. Vials filled with air (20.9% oxy-
provements could stabilize these compounds and shorten their gen) were substituted in some experiments for a 20.0% stan-
development cycle. The implementation of MAP for solid dosage dard. In brief, gas from each respective cylinder was fed slowly
forms is clearly on the horizon. into a portable glove bag (Spilfyter, Thomas Scientific, Swedes-
The key components to successfully using MAP for an oxygen- boro, NJ) by means of a short length of Tygon tubing. After
sensitive product are purging for 30 min to remove residual atmosphere (longer
● packaging that furnishes an effective barrier to oxygen per- purge times provided no benefit), 10-mL clear glass vials (25
meation  54 mm Type I untreated, Kimble/Kontes, Vineland, NJ) each
● incorporation of an inert gas–flushing step that is compa- were filled under a given atmosphere, and 20-mm gray rubber
tible with the speed of current product packaging lines stoppers (S-87J, 4405/50, West Pharmaceutical Services, Li-
● selection of an appropriate inert gas (e.g., argon, nitrogen, or onville, PA) immediately were positioned into the vials. After
carbon dioxide) nearly 50 vials were filled, they were removed from the bag,
● rapid means to measure residual headspace oxygen content and an aluminum flip-top cap was crimped into place on each
in various pharmaceutical packaging configurations that range vial. Using an oxygen–carbon dioxide analyzer (CheckMate
from very low volume (tens to hundreds of microliters) to 9900, PBI Dansenor A/S, Ringstëd, Denmark), vials from each
relatively high volume (milliliters to liters). prepared lot were analyzed immediately and after three months’
Fast, sensitive methods for monitoring headspace oxygen storage in ambient laboratory conditions (19–22 C, 45–55%
levels would greatly benefit the development and use of MAP RH) for changes in relative oxygen levels. Comparisons of the
and facilitate its implementation in packaging R&D, manu- two timepoints confirmed that no oxygen had leaked into the
facturing process development, and final product quality assur- vials. In fact, this vial–closure combination, which is used for
ance. For packaging R&D, robust oxygen measurement meth- commercial parenteral pharmaceutical products, has been
ods would enable real-time screening of the suitability of shown to maintain headspace integrity for at least two years.
packaging systems (all components of a container and closure) Although the previously described sample preparation
for MAP packaging applications. The most obvious example method undoubtedly fails to produce the exact concentrations
is the determination of oxygen ingress kinetics for a packag- of gases in the prepared vials as those provided in the corre-
ing system under both realistic and accelerated storage condi- sponding certified cylinders, the use of these vials for relative
tions. Experimental data could be used to develop predictive comparisons of the capabilities of each technique should be
models for each critical packaging system component to sup- valid. For the purposes of the experiments, the vial contents
port future MAP development programs. (vial-to-vial variability in each lot was found to be very low)
For manufacturing process development, rapid and accurate are understood to be the same as those of the respective fill
oxygen determination in the development feedback loop would tanks. Of course, any attempt at an absolute measurement of
benefit the scale-up and validation of a manufacturing process vial oxygen contents with any given analytical method will be
that implements maximum package throughput per unit time subject to the same technique biases that this report attempts
while achieving and maintaining critical threshold levels for to address. To facilitate comparison, for a given concentration,
oxygen. Tools that provide quality assurance and quality con- vials from the same lot of prepared samples were used for com-
trol information at-, on-, or off-line would ensure that the oxy- parative purposes for all five measurement techniques. For ex-
gen content of the inert atmosphere in the packaging container ample, of the nearly 50 vials prepared at 5.0% oxygen levels,
headspace is below the predetermined specification for the prod- five were reserved for measurement for all five techniques ex-
uct. Such testing also could assess the container-closure integrity amined in the study. Measurement of relative vial-to-vial con-
and could augment or replace current microbial ingress mea- tent variability using the CheckMate 9900 oxygen–carbon diox-
surement methods that are based on sterility assessment. ide analyzer indicated excellent content uniformity (no
The rapid and accurate determination of headspace oxygen differences) among five randomly pulled vials.
levels in the various packaging forms used in the pharmaceu- Instrumentation. Headspace gas chromatography (GC). Headspace GC
tical industry poses several unique analytical challenges that was implemented using a GC system (HP 6890, Agilent Tech-
Pharmaceutical Technology JULY 2002 45
 surements. The instrument was allowed to equi-
librate for 30 min before taking measurements.
Soap-bubble meter The sample holder was purged constantly with
Recorder
dry nitrogen set at a flow rate of 3 standard
L/min, and measurements were acquired at a
Syringe sampling rate of 100 Hz.
Detector Fluorescence quenching. A miniature spectrom-
Flow Electrometer
Two-stage splitter or bridge eter (SF2000, Ocean Optics, Inc., Dunedin, FL)
pressure Rotometer Injector equipped with a 300-m bifurcated optical
regulator fiber and a 21-gauge, custom-designed needle
A/D probe was used for oxygen analysis. The cus-
tom needle probe consisted of a 2-mm shaft
Flow
controller opening offset by 2 mm from the pointed tip
Data of the needle. The blistal end of the aluminum-
Carrier-gas Column
system clad bifurcated optical fiber was fed into the
supply needle and exposed at the shaft opening. The
exposed optical fiber tip was coated with a fluo-
Column oven rophore sol-gel film. Fluorescence intensity as
a function of oxygen concentration was moni-
Figure 1: The components of a typical GC instrument.
tored by means of the same fiber at a charge-
coupled device 2048-element silicon detector
array.
nologies, Wilmington, DE) equipped with a thermal conduc- Quadrupole mass spectrometry. A quadrupole mass spectrometer
tivity detector. Separations were performed using an HP porous equipped with a miniature array of nine quadrupole mass ana-
layer open tubular (PLOT) Molesieve 5-Å column (30 m long lyzers (Micropole, Ferran Scientific, San Diego, CA) was used for
 530 m i.d.) with 50-m film thickness (Agilent). The thin- oxygen determination. The system’s roughing and turbomole-
film HP PLOT Molesieve 5-Å capillary column is a fused- cular pumps achieved low base vacuum pressures (<107 Torr).
silica column coated with a homogeneous layer of 50-m-thick A thoriated iridium electron source ionized the sample before
5-Å molecular-sieve zeolite stationary phase. Because of the quadrupole mass filtering. The ions formed at the source were
column’s sensitivity to moisture, a moisture trap was installed extracted and focused into the entrance apertures of the quadru-
in-line with the carrier-gas feed. Injections were performed pole array by a series of lenses. Detection was achieved at a Fara-
with a 25-L gas-tight locking syringe (Hamilton, Reno, NV) day cup positioned at the opposite end of the quadrupole array.
fitted with a 22S-gauge needle with a #2 point style. Chroma- The instrument was operated in a trend gas–only mode wherein
tographic conditions were as follows: the analysis residence time per each gas monitored was set to
● carrier gas: helium (2 mL/min) 4.5 s, and the partial pressures of each monitored gas were
● oven temperature: 26 C recorded during the course of the measurement. An analog mass
● inlet: 160 C, split mode, 10:1 split ratio, 20 mL/min split flow spectra output that allows for species identification throughout
● injector: 160 C the mass range (2–300 amu) of the detector was used.
● run time: 10 min

● TCD detector: 160 C. Headspace GC

A series of standards and system-suitability samples also was Principles of operation. When a problem arises in a pharmaceu-
analyzed. tical research laboratory that requires analysis of a headspace
Electrochemical methods. The authors used the CheckMate 9900 sample, headspace GC is probably the first method that comes
oxygen–carbon dioxide analyzer equipped with a solid-state to mind for solving the problem. This approach is reasonable
zirconia ion-selective electrode for oxygen determination and given that a vast body of knowledge, stretching back 40 years,
a nondispersive infrared spectrometer for carbon dioxide mea- describes its theory and application (3). Although headspace
surements. The instrument was warmed up for 10 min before GC is an analytical problem-solving tool with capabilities far
measurements were taken and was calibrated according to the beyond the bounds of oxygen analysis, other techniques can
vendor’s specification. The instrument was set to withdraw provide similar or better results for specific headspace analytes,
2 mL of headspace sample using a small internal-diaphragm often in a more timely and simplistic fashion. The following
pump that fed the sample into a small cell that contained the description of headspace GC is intended only as a cursory
measurement electrode. The carbon dioxide measurement mod- overview. The authors refer those who are interested to the sub-
ule was turned off during the experiments. stantive literature about the topic for greater detail (3).
Frequency modulation spectroscopy. A headspace oxygen analyzer GC is inherently a technique for the study of volatile com-
(FMS-760, Lighthouse Instruments, Charlottesville, VA) equipped pounds. Figure 1 is a schematic of the components that con-
with a tunable diode laser source (762 nm) and photodiode de- stitute a typical GC instrument that is equipped with a split
tector was used for frequency modulation spectroscopic mea- injection port and a capillary column. For headspace sampling,
46 Pharmaceutical Technology JULY 2002 www.phar mtech.com
a gaseous sample is introduced by injection into an inert mov- for a given concentration, vials from the same lot of prepared
ing gas stream called the mobile phase or carrier gas. samples were used for comparative purposes for all five tech-
The mobile phase transports the sample onto a stationary niques discussed in this report. The technique displayed good
phase where the components of the sample are separated by linearity throughout the concentration range explored (r2
means of selective partitioning between stationary and mobile 0.9903; data not shown), but a substantial y offset was observed
phases. The time, referred to as retention time (tR), that elapses that was studied in subsequent experiments. A series of experi-
between sample introduction and the point at which peaks ap- ments was performed to examine
pear at the detector is characteristic of the properties of the ana- ● amount of leakage of the package during sample removal

lyte molecules. The area of the respective peaks as they appear ● syringe leakage during sample transport from package to in-

in the resultant chromatogram is proportional to the analyte strument

concentration in the sample, as related by a measurable response ● septum port leakage during sample introduction.

factor. As applied to the analysis of oxygen in pharmaceutical For the type of sample studied, all three sources showed neg-
packaging applications, a sample of headspace (tens of micro- ligible amounts of sample contamination.
liters) is removed from a package by withdrawing a portion of Another source of bias introduction that was examined was
sample with either a gas-tight syringe or a GC autosampler to residual air from the syringe barrel. Three sizes of syringe
pressurize the container with mobile phase to displace an aliquot needles were used, and a direct correlation between bias and
of sample. Performing the experiment in either manner requires syringe-barrel volume was observed. Purging the syringe-
operation in what is referred to as the static mode of headspace needle volume with carrier gas before sampling reduced bias
sampling, wherein the goal is to sample the equilibrium con- from 4689.9 area counts by approximately tenfold, giving 0.08%
tents of a package headspace at a given instance. for the 0.00% oxygen standard. Precision was evaluated by per-
Figure 2, which shows an example chromatogram of lab air, forming five replicate analyses (with a 22S-gauge needle, no
illustrates the determination of headspace oxygen by means of purging) of 20.9% (air, high) and 0.00% standards (low); these
a megabore capillary column with thermal conductivity detec- experiments yielded an average of 22.1% and 0.83% with a
tion. Specific experimental conditions for the analysis are de- relative standard deviation (RSD) of 1.6% and 10%, respec-
scribed in detail in the “Experimental” section of this article. tively. The limit of quantitation (LOQ) (signal to noise [S/N]
As in any type of chromatographic analysis, selection of the ap-
10) of oxygen for the technique was estimated at 0.25%
propriate column to achieve separation of the sample gas mix- without purge and 0.025% with purge step. However, sampling
ture into its various components is pivotal. bias must be understood for any specific system before accu-
To separate oxygen from argon and nitrogen in an air sample rate data at low levels can be realized.
during a 10-min analysis, the authors used a commercially avail- Applications. The use of headspace GC to analyze oxygen in
able thin-film PLOT capillary column consisting of a fused- pharmaceutical packaging configurations offers several ad-
silica capillary tube coated with a homogeneous 50-m-thick vantages such as the widespread availability of instruments and
zeolite molecular sieve (5-Å pore size) film. The mode of sepa- user familiarity. For sample containers of appropriate dimen-
ration is based on the size of the analyte molecule or atom; sions, the autosampling capability of commercial headspace
retention time is as follows: nitrogen > oxygen > argon. The net GCs makes the instrument particularly appealing for analyz-
area and area percentage of oxygen are determined using the ing samples en masse. GC autosampler vial dimensions typi-
following equations cally are not consistent with common pharmaceutical package
sizes, thus the use of an autosampler usually is limited to
 He   x either a few samples of appropriate dimension or samples that
AreaNet, x  AreaRaw, x can be transferred to headspace autosampling vials. The latter
 He
would defeat the purposes of many of the investigations
described in this article.
AreaNet, x Withdrawing and injecting samples manually with a gas-tight
AreaNet, x %  100%
AreaNet, Ar
AreaNet, N
AreaNet, O syringe can pose various problems. For example, when work-
ing with samples that are under vacuum, pressure equalization
with ambient air can lead to sample dilution and erroneously
in which x is argon, nitrogen, or oxygen; AreaRaw,x is raw peak large values for oxygen. Therefore, one must use a gas-tight
area of x (mV s); AreaNet,x is net peak area of x after conversion; locking syringe and institute appropriate controls as described
He is the conductivity of helium (154.6 W/m K); and x is the in the previous discussion. Moreover, one must carefully re-
conductivity of argon (17.8 W/m K), nitrogen (25.9 W/m K), move samples from packages to avoid introducing ambient at-
or oxygen (26.2 W/m K) (4). mospheric leakage during the sampling process. Our experi-
The authors evaluated the linearity of headspace GC for the ence showed that applying a self-adhesive rubber septum before
determination of oxygen in the range of 0.00–20.9%. A gas- puncturing the sample usually was sufficient as a preven-
tight locking syringe was used to extract and manually inject tative measure when puncturing aluminum foil (e.g., foil
25 L from the contents of vials that contained 0.00, 5.0, 10.0, induction–sealed high-density polyethylene [HDPE] bottles).
or 15.0% oxygen, and lab air (see the “Experimental” section Sealing samples with rubber stoppers (e.g., lyo vials) usually
for details about sample preparation). As previously mentioned, provides an adequate barrier to leakage during sampling. Sam-
48 Pharmaceutical Technology JULY 2002 www.phar mtech.com
 O2 REF
E RT ln
240 nF O2 UNK
Nitrogen
200 in which E is potential (V), R is the gas constant (J/K), T
is the temperature (K), n is the number of electrons trans-
Intensity (mV)

160
ferred, and F is the Faraday constant (J/V) (5).
120
Oxygen In practice, the measurement procedure involves punc-
turing the package and removing 2–4 mL of headspace for
80 purging and subsequent analysis in a small cell that con-
tains the electrode. The analysis is rapid, typically requir-
40 ing 5 s. The measurement is not adversely affected by hu-
Argon midity, but the sensor can be damaged if it is inadvertently
exposed to liquid samples. Some commercial instruments
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0
simultaneously measure oxygen and carbon dioxide levels
Time (min)
in headspace samples and are available with sampling acces-
sories for a variety of package containers (e.g., aluminum
Figure 2: An example chromatogram illustrating the determination of
cans, beer bottles). Other commercial electrochemical
headspace oxygen by GC with a PLOT molecular sieve column with thermal
oxygen-analysis techniques are based on the direct reduc-
conductivity detection.
tion of oxygen at a planar gold electrode (Clark techniques
or modifications thereof), but the authors found these tech-
pling errors are common to all destructive headspace oxygen niques to offer no advantage compared with those that are de-
analysis methods, and they demonstrate the difficulties of quan- scribed in this article, thus they are not discussed.
titatively measuring small amounts of an oxygen analyte with- In the present study, a single-point calibration of the instru-
out contaminating it with air. Despite problems with sample ment in laboratory air was performed (yearly calibration typi-
removal, the low absolute sample volume requirements (tens cally is advocated), followed by immediate analysis of a set of
of microliters) make static headspace GC an attractive and po- prepared standards (see the “Experimental” section). The lin-
tentially useful analysis tool for pharmaceutical packaging forms earity of the analytical method was evaluated for oxygen over
of all types and sizes. the range of 0.00–20.9%, and the technique was found to de-
In addition to sample introduction, the major limitations of liver good linearity (r2 0.9984; data not shown). The authors
this technique for the analysis of oxygen are the requirement of evaluated precision by performing five replicate analyses of
a relatively lengthy separation step (a few minutes at best) to be 20.9% (air, high) and 0.00% standards (low). These experiments
oxygen selective, the need to assess appropriate standards, and yielded an average of 20.6% and 0.11% with an RSD of 0.00%
the need to demonstrate system suitability throughout the analy- and 13%, respectively. The vendor-indicated LOQ for the par-
sis of a sample set. Although traditionally a GC instrument has ticular instrument was 0.0001% (1 ppm). As with GC, sampling
a large footprint that would make at-line work unfeasible, the errors may contribute to the observed bias, as discussed later
recent commercial development of portable GC units (some- in this article.
times termed micro-GCs) is beginning to make at-line sam- Applications. The sample-volume requirements of electro-
pling more feasible. chemical methods limit the use of this analysis approach to
large-dimension pharmaceutical packaging types with 2-mL
Electrochemical methods headspace volume such as HDPE bottles, glass vials, and blis-
Principles of operation. The use of electrochemical methods for ter packages (individual blisters) of sufficient volume. The
oxygen analysis also has long historical precedence from both rugged character of such instruments and their ease of use make
a scientific and a commercial perspective. The basis of opera- the technique particularly suited for the manufacturing setting.
tion of most commercial electrochemical headspace-oxygen
analyzers is a solid-state zirconium oxide ion-selective elec- Frequency-modulation spectroscopy (FMS)
trode (5). With this type of sensor, oxygen ions migrate into Principle of operation. Recently commercialized for package in-
defect sites in the ceramic lattice structure of zirconia at ele- spection applications, FMS is a high-sensitivity laser absorp-
vated temperatures (>400 C). As has been known since the tion technique for nondestructive monitoring of gas concen-
time of Nernst (1899), when gases of differing concentrations trations in optically transparent containers (6). Oxygen absorbs
reside on opposite sides of an electrode membrane, a measur- near-infrared (NIR) light in a band of rotational transitions
able potential (E) difference is generated. If one side of the centered at 762 nm. Oxygen absorbance measurements at this
membrane is exposed to gas of reference-oxygen concentra- wavelength with standard spectrometers that use incoherent
tion ([O2]REF), then the potential generated in the presence of white-light sources lack the sensitivity to provide a useful mea-
a gas of unknown oxygen concentration ([O2]UNK) is propor- sure of headspace oxygen levels because of low-frequency lamp
tional to the difference in oxygen concentration across the intensity fluctuations. However, the measurement of S/N can
membrane, as shown by the Nernst Equation be improved vastly (100–10,000 ) with laser-light sources and
50 Pharmaceutical Technology JULY 2002 www.phar mtech.com
 cies, c  . When the laser is scanned through
(a) the wavelength of oxygen absorbance, the amount
of light absorption, which is proportional to the
Recorder
c   c  
gas concentration, is “written” into the side-band
c c c
frequencies by recording the difference in absorp-
c
 c

tion between the two side bands. The differential
(b) absorption information is recovered by means of
Current and
Diode phase-sensitive detection techniques. Figure 4
temperature Sample Detector
controller
laser shows the demodulated absorption line shape. The
amount of light that is absorbed is directly pro-
portional to oxygen concentration. The gas den-
Mixer
Scan rf Lo RF sity, n, is related to the peak-to-peak signal am-
Amp
controller oscillator plitude, I, according to Beer’s law, which for a
IF Data weakly absorbing molecule is shown by
system
••
n cm3
A/D converter I o•S •x
Computer
in which  is the change in intensity of light after
passing through the container (W/cm2),  is the
full width at half maximum of the absorption sig-
Figure 3: a) Frequency and intensity profile of a diode laser beam, from left to right:
nal (cm1),  is a constant, Io is the incident laser
unmodulated, modulated with no absorption, and modulated with absorption by the
intensity (W/cm2), S is the integrated absorption
upper sideband. b) Schematic diagram of an instrument for frequency modulation
cross section (cm2  cm1), and x is the container
spectroscopy. The frequency modulated diode laser output is converted to an
diameter (cm). The measured density in a sample
amplitude modulation after passing through a gas sample which absorbs at a
vial is referenced to a standard and displayed as a
particular wavelength. The amplitude modulation is proportional to gas concentration
concentration in percentage.
and can be phase sensitively detected and related to oxygen concentration.
The authors investigated the capability of the
technique to measure oxygen in 10-mL clear glass vials. A
0.4 single-point calibration of a factory-supplied 20% oxygen stan-
0.2
dard was used to calibrate the instrument. The linearity of the
20% analytical approach was evaluated for oxygen over the range
Instrument response (V)

0.0
of 0.00–20.0% by analysis of 0.00, 5.0, 10.0, 15.0, and 20.0%
0.2
standards. Each sample was measured by placing a vial in a
0.4
13% small optical sample holder and collecting absorption data for
0.6
each sample for 10 s at a sampling frequency of 100 Hz. The
0.8 8%
technique displayed good linearity in the range of concentra-
1.0 4% tions examined in the study (r2 0.9948; data not shown).
1.2 2% Precision for the technique was evaluated by performing five
1.4 1%
replicate analyses of 20.0% (high) and 0.00% (low) standards.
0%
1.6
These experiments yielded an average of 19.8% and 0.19% with
Wavelength (arbitrary units)
an RSD of 0.05% and 26%, respectively. Because the instru-
Figure 4: Frequency-modulation signals from oxygen absorption. The ment is nondestructive, the precision of replicate analyses on
peak-to-peak amplitude of each spectra is proportional to oxygen the same vial also was evaluated by measuring, removing, re-
concentration (noted to the right of each scan). inserting, and remeasuring the sample. Five repeat measure-
ments on a single 20.0% and 0.00% standard indicated good
precision for the method, with an RSD of 1% and 34%, respec-
frequency modulation detection techniques. When a tunable tively. The vendor-specified quantification limits for the mea-
diode laser is modulated with a radio frequency oscillator, the surement were estimated at 0.38% oxygen with a S/N of 2 and
detection bandwidth can be shifted to high frequency at which 1.9% oxygen with a S/N of 10.
intensity fluctuations, inherent to low-frequency measurements, Applications. The application of this technique is limited to
are minimized. headspace monitoring in packages that are optically transpar-
The measurement is conducted by frequency modulating a ent (e.g., glass vials, ampuls, and bottles, including colored glass-
diode laser by superimposing a radio frequency oscillation, , ware such as amber) or translucent (low-density polyeythylene
onto the diode-injection current. The spectral output of a but not HDPE). Other containers with various optical densi-
frequency-modulated diode laser, shown at the top of Figure ties and corresponding transmittance properties also can be
3, consists of a carrier frequency, c, and side-band frequen- sampled but must be evaluated on a case-by-case basis. Pack-
52 Pharmaceutical Technology JULY 2002 www.phar mtech.com
 Quenchers act by depleting the excited-state population (M*),
thus lowering the intensity of luminescent emission or shorten-
20 mA blue— ex 470 Bifurcated
green LED optical fiber
ing decay time. In the case of commercial oxygen sensors, M is
max 470 nm a fluorescent ruthenium dye whose fluorescence intensity is
Linear quenched by oxygen in a manner commonly referred to as dy-
CCD-array namic fluorescence quenching. In terms of the process previously
silicon detector described, a collision of an oxygen molecule (Q) with the fluo-
rophore in its excited state (M*) leads to a nonradiative energy
transfer and a loss of fluorescence intensity.
em 600 nm The amount of fluorescence-intensity quenching can be quan-
Thin sol-gel
film containing titatively related to the partial pressure of oxygen in a sample
Ru complex from the simplified Stern–Volmer equation (8)
Io
Figure 5: Schematic diagram of a headspace oxygen analyzer based
1
K pO2
I
on fluorescence quenching.
in which I is the quenched fluorescence signal intensity, Io is the
ages are inserted into a sample holder that can be customized unquenched fluorescence signal intensity, K is the quenching
for cell volumes of 1–100 mL, which typically present cell constant that relates to a particular fluorophore used as M, and
path lengths of 1–3 cm. pO2 is the oxygen partial pressure.
The method’s principal advantage is that it can nondestruc- An expanded version of the Stern–Volmer equation con-
tively analyze oxygen contents within sealed containers without taining a factorial expansion of these same terms can also be
changing the sample. Thus, a single package can be used to study used to relate Io to oxygen levels in a more exact manner. Stan-
oxygen concentration profiles at periodic intervals and in various dard operation includes, at minimum, a daily calibration at sev-
storage conditions without the need to destructively insert a probe eral oxygen concentrations at a fixed pressure and temperature
or remove the headspace contents for analysis. The value of this and use of the measured calibration coefficients to determine
advantage cannot be overstated in terms of absolute sample re- the concentration of oxygen in an unknown sample.
quirements, sample preparation time, and sample-to-sample varia- In the current study, an instrument that is based on
bility. From a quality assurance perspective, substantial cost sav- fluorescence-quenching technology was calibrated with six oxy-
ings can also be realized from a nondestructive analysis approach. gen concentrations (see “Experimental” section) before initiat-
ing analysis. Although a single-point calibration is recom-
Fluorescence-quenching methods mended, the authors have found that this practice is insufficient
Principles of operation. Oxygen-sensing technologies that are based to provide reasonable accuracy over a 0.00–20.9% oxygen con-
on luminescence quenching are also a relatively new com- centration range. Moreover, a second-order polynomial fit of
mercial development, and the scientific underpinnings for their the data is required to generate calibration coefficients across
operation are described extensively in a recent review article (7). wide ranges of sample concentrations (more than a few per-
This discussion will introduce briefly this oxygen measurement cent). Also, frequent recalibration is necessary because the sen-
technology; the reader is referred to more-specialized literature sors tend to degrade under normal operating conditions. This
for further detail. Figure 5 shows a commercial configuration in characteristic is particularly true of new sensors, which may re-
which fluorescence is used to measure the partial pressure of quire “burn in” for a few hours before stable operation is
oxygen by using a bifurcated optical fiber to transmit an excita- achieved. Analyses were accomplished by inserting a custom
tion source from a blue–green LED (max 470 nm) to a thin- needle probe (see “Experimental” section) into a 10-mL sam-
film coating applied to the tip of the fiber. A ruthenium com- ple vial and waiting 1 min for equilibration of the sensor to the
plex dispersed in the thin film is excited and produces a headspace environment. Shorter equilibration times can be used
fluorescence emission maximum at 600 nm that is collected at and are a function of the chemical composition of the protec-
the tip and carried back to a charge-coupled device detector. The tive overcoat applied to the sensor tip to exclude ambient light
intensity of fluorescence observed is inversely proportional to (a PTFE coating was used in this study).
the amount of oxygen in a liquid or gaseous sample. The linearity of the analytical approach was evaluated for
Luminescence-quenching sensors can be explained by the oxygen over a 0.00–20.9% range in the present study. The
following processes in which M represents the luminescing fluorescence-quenching technique demonstrated a linear rela-
molecule and Q is a quencher: tion between measured versus standard concentration (r2
0.9952; data not shown), but an offset in the data predicted
Photon absorption (Ia) M
h M* 0.83% oxygen for the 0.00% standard, which measured at
Luminescence (kr) M* M
h 1.6%. The accuracy of fluorescense quenching is commonly
known to be limited by resolution (random noise), deviations
Nonradiative decay (knr) M* M

from the Stern–Volmer relationship, and calibration error (8).
Dynamic quenching (kq) M*
Q M
Q* Which, if any, of these factors may have promoted inaccuracy
in the previously described experiments has yet to be experi-
54 Pharmaceutical Technology JULY 2002 www.phar mtech.com
 rience in our lab also has indicated a need to consistently use
longer signal-integration times during the life of any given probe
to achieve sufficient signal intensities (1800 counts over the mea-
Ion with unstable trajectory surement time at zero oxygen) over background noise. The de-
creased response factor that was observed could be the result of
changes in the ruthenium dye (e.g., degradation) over time.
Ion
Higher signal intensities or lower noise levels are desirable be-
collector
cause S/N is proportional to the square root of integration time.
Ion with Moreover, as can be seen from the Stern–Volmer equation, the
stable trajectory rate of change of signal intensity with quenching is highest at
lower oxygen levels. Further development of the instrumenta-
dc and rf tion for oxygen-specific measurements would benefit greatly the
Ionizing voltages overall utility of the technique.
electron beam
Quadrupole mass spectrometric analyzers
Figure 6: Block schematic of a quadrupole mass spectrometer. Principles of operation. Quadrupole mass analyzers are by far the
most common type of mass spectrometer in use today, and the
literature about this type of analyzer is extensive (8). Quadru-
mentally isolated. Measurement precision was evaluated by per- pole mass analyzers often are thought of as mass filters because
forming five replicate analyses of 20.9% (air, high) and 0.00% they can be tuned to transmit ions of a narrow range of mass-
(low) standards; these experiments yielded an average of 20.3% to-charge (m/z) ratios. Figure 6 shows a generalized block
and 1.6% with an RSD of 1% and 17%, respectively. schematic of a quadrupole mass spectrometer. A typical quadru-
Applications. Fluorescence-quenching sensors do not actually pole instrument separates ions with differing masses by apply-
consume headspace oxygen and thus can be miniaturized for ing a combination of static- and radio-frequency electric fields
use on very small volume samples. Still, the technique is de- to four cylindrical rods. A headspace gas sample is introduced
structive because the tip of a 300-m probe must be used to at an inlet and fed into an ion source where electrons are emit-
puncture the package and come in contact with the relevant ted from a filament and ionize the sample gas. The sample ions
headspace, allowing an effective smallest volume of 100 L then are accelerated in an electrical field and are injected into
to be probed. the opening at the center of the rods. In the most simple sys-
Some attributes of the technique include tems, one pair of rods is connected and attached to the positive
● rapid response times (from 5 s to 2 min) that are limited to terminal of a dc power source, and the other connected pair is
the time required for equilibrium to be reached between oxy- attached to the negative end of the same source. A variable rf
gen in the sample headspace and the sensing film by way of field with phase differences of 180 also is applied to each con-
diffusion through the protective overcoat nected pair of rods with an ac power supply. For an ion to travel
● very low sample volume requirements (100 L) that are through the quadrupole to the detector, the ion must have a
limited only by the contact dimensions of the probe stable trajectory in the presence of these applied voltages and
● low cost and small instrument footprint. electric field, as described by the Mathieu equations of motion
The low sample volume requirement makes the fluorescence- for ions. The application of a particular set of voltages to the
quenching sensor the only instrument besides headspace GC rods allows mass filtering and the transmission of a band of ions
that is capable of rapidly measuring oxygen levels in individual with a limited range of m/z values. This process usually is accom-
blister cavities of typical dimensions used in pharmaceutical plished by scanning the rf voltage and dc voltages from zero to
packaging. In addition, the sensor’s small size and easy setup a predetermined maximum value while maintaining a fixed ratio.
make it very attractive from a portability perspective for achiev- In this manner, a mass spectrum of analyte ions reaching the
ing at-line packaging measurements. Fluorescence-quenching detector can be recorded at a Faraday-cup detector.
instruments are also available in multichannel configurations The quadrupole mass analyzer used in this study comprised
that allow the simultaneous measurement of several samples, a set of 16 identical cylindrical rods arranged to form nine
and the data-logging component of the software executes pro- quadrupoles (8). The mass resolution of the quadrupole was
grammed measurements at fixed time intervals. 1 amu, and the mass range of the analyzer was 2–300 amu.
The fluorescence-quenching method does have some limita- The sample was introduced to the instrument by means of a
tions, however. Care must be taken to calibrate throughout the custom-made analysis port with a toggle switch to introduce
intended measurement temperature range of samples because sample from a needle used to puncture a sample package. How-
fluorescence-quenching levels change as a function of tempera- ever, other package-puncturing sampling techniques could also
ture. Temperature affects the measurement by modifying fluo- be used (e.g., syringe and injection port). The linearity of the
rescence decay time and the collisional frequency of oxygen with analytical approach was demonstrated for oxygen over the range
the fluorophore. Requirements to maintain the sample at 1 C of 0.00–20.9% (r2 0.9965; data not shown). Precision was
or to concurrently measure temperature while conducting the evaluated by performing five replicate analyses of 20.9% (air,
measurement curtail the practical utility of the method. Expe- high) and 0.00% (low) standards; these experiments yielded an
56 Pharmaceutical Technology JULY 2002 www.phar mtech.com
 QA/QC setting other con-
Table I: Performance parameters for the oxygen-analysis methods evaluated. siderations such as long-
Standard Measured Precision Accuracy Linearity LOQ term method robustness,
Method (%)* (%)** Std. Dev.† RSD (%) Bias ‡ r2 (%) § stability, ease of calibration
Headspace GC 20.9 22.1 0.30 1.6 1.20 0.9903 0.25 and use, oxygen levels to be
0.00 0.83 0.08 10.0 0.83 routinely analyzed, and
Electrochemical 20.9 20.6 0.00 0.0 0.30 0.9984 0.0001** other practical considera-
0.00 0.11 0.01 13.0 0.11 tions will help guide tool se-
lection. The forthcoming
FMS 20.0 19.8 0.01 0.05 0.20 0.9948 1.9
discussion will examine tool
0.00 0.19 0.05 26.0 0.19
selection criteria in terms of
Fluorescence 20.9 20.3 0.20 1.0 0.6 0.9952 unknown performance considerations,
quenching 0.00 1.6 0.27 17.0 1.6 other relevant technique at-
Quadrupole MS 20.9 21.9 0.10 0.50 1.0 0.9965 sub-ppm tributes, and type of pack-
0.00 0.00 7.54  105 2.60 2.9  103 age being considered.
* Prepared as described in the “Experimental” section of this article. First, what can we learn
** Average of five different samples. about how the five tech-
†
Standard deviation. niques presented in this ar-
‡
Bias calculated as measured—theoretical. ticle performed in terms of
§
Limit of quantitation, S/N
10, all specified by instrument supplier with the exception of headspace
GC, which was measured.
measuring oxygen concen-
trations in the range of zero
to roughly ambient oxygen
average of 21.9% and 0.00% with an RSD of 0.5% and 2.6%, levels? The data shown in Table I are clarified in the following
respectively. The very small amount of offset observed for this discussion of the performance parameters that were studied.
instrument was the result of the extremely low LOQ (estimated Key points. Linearity. All five techniques that were investigated
at sub-ppm levels) of the measurement technique, limited only displayed good linearity over the measurement range
by the lower limit of vacuum achieved in the analyzer. 0.00–20.0% or 0.00–20.9%, all displaying r2 0.9903.
Applications. The application of quadrupole mass analyzers Measurement precision. The standard deviation and RSD at both
in a residual gas–analysis mode has long precedence in the high (20.0%) and low (0.00%) ends were examined for each
semiconductor industry in which residual contaminant gas technique. As the data show, the precision for the techniques at
levels in various ultrahigh vacuum chambers must be moni- low concentrations ranged from 2.6% to 26%. Note that the
tored and controlled. The principal advantage of using quadru- RSD of five replicate measurements at ambient oxygen levels
pole mass spectrometry for pharmaceutical packaging appli- were 2% for all five of the techniques.
cations is that many analytes can be monitored simultaneously Accuracy. Generally speaking, for most MAP programs the aim
and quantitated (2–300 m/z) with one injection of headspace is to reduce oxygen values from ambient to a predetermined
sample. As is the case for other destructive techniques, the specified level. Often, the measurement technique may not be
largest drawback for this instrumentation is sample introduc- required to give the absolute oxygen level in the container but
tion. Several sample-introduction strategies are being explored only to be sufficiently accurate to indicate that the oxygen con-
to improve the utility of the approach. The technique likely tent of the package is below the target threshold, thus serving
would be useful in process-monitoring applications in which as a limit test. Thus, the importance of the measurement tech-
the detailed contents of an inert gas blanket beyond that of nique to measure at low levels (e.g., 5%) is more important
simple oxygen levels is required. than capabilities at high levels. One way to interpret the per-
formance data in Table I in this context is to examine the ab-
Which technique should I use? solute differences observed between theory and experiment for
The answer to this important question depends on the type of the low-level oxygen standard. As Table I shows, the degree of
information desired, instrumentation availability, and the type deviation (from least to greatest) that was observed between
of packaging being considered. In MAP development, the stud- predicted and measured low oxygen level standards followed
ies may comprise research into the oxygen permeability prop- this trend: quadrupole MS < electrochemical < FMS < GC <
erties of various packaging systems, packaging-system selec- fluorescence quenching.
tion, process development, process implementation, and finally, LOQ. The lower LOQs shown in Table I were not measured di-
quality assurance. For the evaluation of various packaging sys- rectly as part of the study. In instances in which they are pro-
tems, the most common questions may be about understand- vided, the values are either estimated or obtained from the ven-
ing the rates of oxygen ingress (or consumption) ranging from dor. For critical absolute oxygen concentration measurements
very low levels (1% oxygen) to ambient oxygen concentra- at very low levels, the lower LOQs should be established for the
tions. The types of tools used in a research laboratory should specific system being studied. Clearly, the sampling method
be reevaluated when considering a manufacturing setting for plays a critical role in the quality of the data for oxygen levels
process development and implementation. In addition, in the 2% and limits the accuracy of the data.
58 Pharmaceutical Technology JULY 2002 www.phar mtech.com
 Table II: Some attributes of various headspace oxygen–sampling tools.
Analysis Ease of Instrument Required
Instrument Time Calibration/Use* Availability Expense Sample Volume Destructive
GC Minutes 3 Commercial $40K
10–30 L Yes
Electrochemical Seconds 1 Commercial $4–10K 1–3 mL Yes
Fluorescence
quenching Minutes 3 Commercial $2–5K None** Yes
FMS Seconds 2 Semicommercial $35–50K None† No
Quadrupole MS Minutes 3 Semicommercial $25–50K 10–30 L Yes
* 1 easiest to calibrate and use; 2 more difficult to calibrate and use; 3 most difficult to calibrate and use.
** A sufficient sample volume must be available so that a probe can be inserted and make contact with the
sample, estimated at 100 L.
†
Instrument is nondestructive and can be used on vials through which a laser can be transmitted.

Knowing the performance limitations of a particular in- and they display an unparalleled ease of use. These instruments
strument is key to interpreting the results obtained from a se- also tend to be among the most rugged of all the instrumen-
ries of measurements. Other considerations also affect tech- tation described herein and seem to be the most well suited for
nique selection. Table II summarizes some additional points use by operators with a limited amount of technical training.
to consider when matching an analytical requirement to a head- Glass vials and ampuls. Glass vials and ampuls are another com-
space oxygen–analysis tool. The table shows that analysis times monly used pharmaceutical packaging form, especially for paren-
for the five techniques range from minutes (GC, fluorescence teral formulations. Any of the five techniques discussed in this
quenching, quadrupole MS) to seconds (electrochemical meth- report can be used to analyze headspace oxygen levels as long
ods, FMS). All the instruments are available commercially, al- as sufficient volume is available. FMS offers considerable advan-
though at various stages of development, and vary in expense tages for this package type because it is nondestructive. In ad-
from $2,000 to $50,000. FMS and quadrupole MS are the least dition, FMS can be used for vials filled with a variety of media,
commercially developed of the techniques at this time, and including powders that may coat the walls of the container. As
more work is needed before their routine use will be possible. long as a portion of the analysis beam is transmitted, the dif-
Ease of calibration and use, which is a highly subjective cate- ferential absorptive signal processing inherent to the technique
gorization, reflects the authors’ judgments about the relative can be used to yield oxygen concentration values.
ease of training others to operate the equipment. Other fac- Blister packages. The small volume of individual blister cavities
tors such as versatility, portability, and suitability for use in a makes this type of packaging the most challenging to analyze
manufacturing setting also are important when one is choos- for headspace oxygen levels. The challenges are twofold: the ex-
ing an instrument. For example, both headspace GC and tremely small volume available for sampling as well as sample
quadrupole MS have numerous uses beyond oxygen monitor- removal or introduction without contamination. All the instru-
ing, and one may want to acquire other information simulta- ments that are applicable for small-volume measurements (i.e.,
neously (e.g., amount of moisture, carbon dioxide, or other headspace GC, fluorescence quenching, quadrupole MS) have
volatiles). FMS is the only nondestructive technique of which drawbacks from a sampling perspective. Although future de-
the authors currently are aware, and it can be used only with velopment in this area is needed, headspace GC seems to be
package types that allow adequate transmittance of the instru- most suitable at this time because self-adhesive septa can be
ment probe beam. placed on the package backing and a small portion of sample
Package types. Identifying the type of package from which can be withdrawn manually for injection. Other viable ap-
oxygen levels must be sampled and analyzed is another con- proaches, including nondestructive techniques, currently are
sideration for determining the appropriate analysis tool. Mak- being investigated.
ing a decision from this angle requires information from both
Tables I and II, especially the sample-volume requirements for Summary
each technique that are listed in the latter. For the sake of brevity, This article has provided an introduction to the principles of
only the three most common pharmaceutical packaging con- five measurement techniques for analyzing the headspace oxy-
figurations (plastic bottles, glass vials–ampuls, and blister pack- gen levels in common types of pharmaceutical packages. Vari-
ages) are discussed. ous performance criteria were applied to examine the fit of each
Plastic bottles with foil-induction inner seal. Plastic (e.g., HDPE) bot- particular tool to typical pharmaceutical package types. Future
tles with foil-induction inner seals (the most common have research about the topic is needed to better address the ro-
33-, 75-, 120-mL volumes) are used commonly to package bustness of sample introduction techniques for small-volume
pharmaceutical products. For the rapid and accurate determi- samples and perhaps to develop sensor technologies that would
nation of headspace oxygen in a package of this type, all the allow oxygen levels to be remotely sampled and profiled for a
techniques discussed in this report apply, with the exception period of time from a small device integrated or inserted into
of FMS. Based on the authors’ experience, the electrochemical the package.
methods are best suited for large-volume (
2 mL) packages,
60 Pharmaceutical Technology JULY 2002 www.phar mtech.com
Acknowledgments 7. J.N. Demas, B.A. DeGraff, and P.A. Coleman, “Oxygen Sensors Based
on Luminescence Quenching,”Anal. Chem. 71 (23), 793–801 (1999).
The authors thank the instrumentation vendors whose coop-
8. (a) M.J. Drinkwine and D. Lichtman, “Quadrupole Mass Analyzers,”
eration greatly facilitated the work described in this article. in Partial-Pressure Analyzers and Analysis, N.R. Whetten and R. Long,
Particularly, special thanks goes to Jim Veale and Bobby An- Jr., Eds. (American Vacuum Society, New York, NY, 1978), pp. 235–278.
derson of Lighthouse Instruments, Inc. The authors also thank (b) R.J. Ferran and S. Boumsellek, “High-Pressure Effects in Minia-
Denise Farquharson for assistance with the preparation of gas ture Arrays of Quadrupole Analyzers for Residual Gas Analysis from
109–102 Torr,” S. J.Vac. Sci. Technol. A 14, 1258–1265 (1996). PT
standards.

Disclaimer
FYI
The information described in this report does not constitute
endorsement by Merck & Co., Inc. of any particular instrument Postdoctoral fellowships
type or supplier. The College of Pharmacy at Rutgers offers postdoctoral fellowships
in partnership with various pharmaceutical companies.
References The fellowships include practical industry experience; an adjunct
1. D.M. Johnson and L.C. Gu, “Autoxidation and Antioxidants,” in The faculty appointment at the College of Pharmacy;and benefits such as
Encyclopedia of Pharmaceutical Technology, Vol. 1, J. Swarbric and J.C. a stipend,health insurance, and vacation and sick days.Applicants
Boylan, Eds. (Marcel Dekker, New York, NY, 1988), pp. 415–450.
must possess a PharmD from an American Council on
2. K.A. Connors, G.L. Amidon, and L. Kennon, Chemical Stability of Phar-
maceuticals (Wiley & Sons, New York, NY, 1979), pp. 80–98. Pharmaceutical Education–accredited college of pharmacy and must
3. B. Kolb and L.S. Ettre, Static Headspace Gas Chromatography: Theory submit a curriculum vitae,a written career-objective statement,and
and Practice (Wiley & Sons, New York, NY, 1997). three letters of recommendation.
4. CRC Handbook of Chemistry and Physics, D.R. Lide, Ed. (CRC Press, For more information or to apply,contact Joseph A.Barone,PhD,
Boca Raton, FL, 72d ed., 1991), p 2040.
Rutgers,The State University of New Jersey,College of Pharmacy,
5. J. Wang, Analytical Electrochemistry (Wiley & Sons, New York, NY,
2000). Department of Pharmacy Practice and Administration,Attention:
6. (a) G.C. Bjorklund, Opt. Lett. 5 (1), 15–17 (1980). (b) G.R. Janik, C.B. Fellowship Program,160 Frelinghuysen Rd.,Piscataway,NJ 08854-
Carlisle, and T. F. Gallagher, J.Opt. Soc. Am. B 3 (8), 1070–1074 (1986). 8020,tel.732.445.2675 ext.600,fax 732.445.5767,
(c) J.A. Silver, Appl. Opt. 31 (6), 707–717 (1992). (d) D.E. Cooper and http://pharmacy.rutgers.edu.
R.U. Martinelli, Laser Focus World 28 (11), 133–146 (November 1992).

Circle/eINFO 57
Pharmaceutical Technology JULY 2002 61