Return to Cover of Wave of the Future
Thomas B. Gold1,2, Robert G. Buice, Jr., Robert A. Lodder1,
George A. Digenis1,*
1Division of Medicinal Chemistry and Pharmaceutics, College
of Pharmacy, University of Kentucky, Lexington, KY 40536-0082.
2Present address: Banner Pharmacaps, Inc. 4125 Premier Drive, High Point, NC
27265.
ABSTRACT
The migration of formaldehyde from a polyethylene glycol (PEG) fill into
the gelatin shell of a soft elastic capsule was monitored using near-infrared
(NIR) spectrophotometry. Principal component regression (PCR) was employed
to analyze the spectra of the empty capsules. Good correlation was established
(r2 = 0. 988) when actual concentrations of formaldehyde in the
PEG fill of the capsules were regressed against the PC values from NIR spectra
of the emptied and washed capsules.
KEY WORDS: Gelatin; Crosslinking; Formaldehyde; Soft Gelatin Capsules;
Near-Infrared Spectrophotometry
*Author to whom correspondence should be addressed.
INTRODUCTION
Gelatin capsules arguably represent the most versatile pharmaceutical
oral delivery system, into which drug powders, solutions, or suspensions
may be filled. Other applications that utilize the hard gelatin capsule
(HGC), soft elastic gelatin capsule (SEGC), or gelatin coating include controlled
release technologies, cosmetic formulations, and nutritional supplements.
The favorable properties of the gelatin capsule include its strong yet flexible
backbone, glossy appearance, ability to hold dyes, and ease of swallowing
(1).
The proclivity of the gelatin molecule toward formaldehyde-induced derivatization
and crosslinking through its amino and guanidino functionalities has been
well characterized (2-6). The formation of crosslinks between these tertiary
functionalities of the polypeptide leads to an increase in the molecular
weight of gelatin and more specifically, a detrimental effect on the dissolution
of the drug which it encapsulates (7). The potential effects of formaldehyde
crosslinking in gelatin have been investigated because of apparent dissolution
problems in in vitro testing of hard and soft gelatin capsules (1,8). Bottom
et al. have shown that SEGCs, containing 20 or 80 ppm formaldehyde in a
lipophilic fill, demonstrate reduced dissolution rates of acetaminophen
compared to fresh, unadulterated SEGCs (9). Using HGCs exposed to atmospheric
formaldehyde concentrations of 150 ppb for as few as two hours, Gold et
al. observed partial insolubilization of the gelatin shell and decreased
dissolution rates of amoxicillin when compared to unstressed capsules (10).
These crosslinking phenomena have been known to effect the formation of
a thin, water-insoluble membrane (pellicle) around the gelatin capsule during
dissolution testing (8). The pellicle thus acts as a barrier that restricts
the release of drug. Recently, much attention has been paid to the source
of the low molecular weight aldehydes implicated in gelatin capsule insolubilization.
It has been reported, for example, that corn starch, a common drug excipient
and fill material in HGCs, may contain low levels of hexamethylenetetramine
stabilizer (1, 11), which decomposes under humid conditions to give ammonia
and formaldehyde. Polyethylene glycols, frequently used as bases for hydrophobic
drugs in soft elastic gelatin capsules (SEGs), liberate low molecular weight
aldehydes through free radical reactions and upon exposure to aerobic conditions
(11,12).
Recently, we have successfully employed near-infrared (NIR) spectrophotometry for determination of water uptake (13,14) and extent of crosslinking in HGCs (10,13). In the latter study, empty gelatin capsules (sans excipients or drug powders) were exposed to formaldehyde vapor. The capsules were then filled with fresh amoxicillin and their dissolution performance was determined in vitro and also successfully predicted using NIR spectrophotometry
(10,13). It was evident to us, however, that a nondestructive and noninvasive
method was needed to detect formaldehyde that emanated from the encapsulated
excipient and migrated into the gelatin shell, with the potential to cause
insolubilization of the latter.
The present work combines the analytical power of NIR spectrophotometry
and a multivariate analysis tool, principal component regression (PCR)(15).
These two methods were utilized to predict, in an empty SEGC, the concentration
of formaldehyde in a polyethylene glycol base originally contained within
the same capsule.
MATERIALS AND METHODS
Materials. Size 8, clear oblong capsules with 0.5 em twist-off tips were
supplied by Banner Pharmacaps, Inc. (Chatsworth, CA). Low-aldehyde polyethylene
glycol (PEG) 400 was obtained from Union Carbide (Danbury, CT). Formaldehyde,
37% w/w in water, was purchased from Aldrich (St. Paul, MN).
Instrumentation. Spectral data were collected by an InfraAlyzer 500 (Bran+Luebbe,
Elmsford, N.Y.) spectrophotometer. Absorbance values were obtained as log
(1/R) data from capsules every 4 run from 1100 to 2500 run using a 10 run
bandpass. The data were collected and analyzed with a PS/2 model 50 and
an HP1000 (Hewlett Packard, Palo Alto, CA) computer, respectively. Analytical
software was written in Mathematica (Wolfram Research, Inc., Champaign,
IL).
Procedure. Capsules were injected with one of five solutions listed in
Table I. Injection (0.5 ml) was accomplished by
puncturing the capsule at the tip of the twist-off tube using a l ml syringe
equipped with a 27 gauge needle. The capsules were wiped at the point of
injection to assure that the solution would not migrate to the outer surfaces
of the capsule. The capsules were placed tip upwards inside a sealed glass
jar for 48 h. Subsequently, the tips of the SEGCs were cut off, with care
taken not to allow the solution inside to touch the outer capsule surfaces.
Using a small pipet, the solutions inside the SEGCs were withdrawn, and
the inside surfaces of the capsules were rinsed two times with fresh PEG
400 and twice with diethyl ether. The capsules were then stored in a nitrogen
atmosphere for twenty four hours. Capsules were scanned in triplicate with
the incised end downward in a 900 conical reflector machined from aluminum
(16). A steel wire support was used to achieve upright vertical capsule
positioning within the cone (10).
RESULTS AND DISCUSSION
Figure 1 shows the NIR spectra for all six experiment
groups of empty SEGCs. Because there were three NIR scans of each sample,
there were three reflectance values at each recorded wavelength. Spectral
averaging, which retained only the average reflectance value at each wavelength,
was performed on the data. The spectra were subsequently smoothed with a
cubic spline algorithm, after which spectral scattering effects, due to
sample inhomogeneity (variation in capsule wall thickness and composition),
were eliminated by multiplicative scatter correction. Non-uniformity of
samples prevented simple analysis techniques from not only explaining the
spectral signal of an individual capsule, but also from differentiating
its NIR spectra from that of other capsules. Thus, principal component regression
(PCR) was used to analyze the smoothed, scatter-corrected data. PCR transforms
a large number of correlated variables into a new set of uncorrelated variables,
called principal components (15). Most of the variation contained in the
original variables is retained by the first few PCs. Multiple linear regression
was utilized to determine which of the PCs correlated to the amounts of
formaldehyde that was injected into the gelatin capsules. Cross-validation
was employed on the entire group of capsules with thirteen PCs. Nine of
the thirteen PCs were found to be of statistical significance (t statistic>
3.0) and to pass cross-validation. Figure 2 illustrates
the strong linear correlation (r2=0.988) of actual concentration
of formaldehyde (solution) in the PEG solutions, originally encapsulated
by the SEGCs, to concentrations predicted by the NIR spectra of the same
capsules, emptied. From cross-validation, the standard error of estimate
(SEE) and standard error of prediction (SEP) were 0.024 and 0.040 v/v%,
respectively, with p S 0.05 from f-test. The calibration line, representing
a theoretical correlation coefficient of r2= 1, is shown superimposed
over the data points (Figure 2). In effect, the changes in the NIR spectra
of the empty gelatin capsules, resulting from formaldehyde migration into
and possible reaction with the gelatin shell, were correlated to a known
concentration of formaldehyde placed in the fill material of the SEGC. With
respect to the interdependence between empty SEGC NIR spectra and known
concentration of formaldehyde doped into the capsules' PEG fill, two possibilities
exist which can explain the rather strong correlation: (1) all, or, (2)
a fixed percentage of the formaldehyde which was originally encapsulated
by the SEGCs in each of the experimental groups (Table I) migrated into
and possibly reacted with the gelatin capsule shell. The latter phenomena
was manifested by concomitant changes in the NIR spectra of the emptied
SEGCs. If, for example, the formaldehyde in the fill of the second and subsequent
groups of capsules (formaldehyde solution, 0.10, 0.20, 0.40% v/v in PEG,
Table I) had not completely migrated into the gelatin shell of the SEGC,
while the formaldehyde in the fill of the first group of capsules (formaldehyde
solution, 0.05% v/v in PEG, Table I) not only migrated into but also reacted
with the gelatin capsule shell, the NIR-predicted value of formaldehyde
concentration in the former groups would have been significantly lower than
the actual amount of formaldehyde incorporated into the capsule fill. The
resulting graph correlating actual concentrations of formaldehyde in PEG,
encapsulated for 48 h in soft gelatin capsules, to concentrations predicted
by NIR spectrophotometry, would have contained points which deviated downward
from the calibration line in Figure 2.
Examination of the transformation matrices connecting wavelength and
PC hyperspace can determine which wavelengths (and possible chemical interactions)
correlate positively (or negatively) with the NIR spectral changes of the
empty SEGCs, originally filled with formaldehyde-tainted PEG. The loadings
for the first PC (Figure 3) showed a positive weighting
on wavelengths between 1100 and 1400 nm and above 2050 nm. Conversely, the
same loadings (Figure 3) demonstrated negative weightings on wavelengths
between 1400 nm and 2050 nm. The second PC loadings graph showed more specific
wavelengths of correlation (Figure 4). For example,
strong positive weighting was established on 1385, 1840, and 2200 run, while
strong negative weighting was placed on 2040, 1945, and 1420 run. Of significance
are the latter two wavelengths: 1945 and 1420 nm, which represent the water
molecule's wavelengths of NIR absorbance. Because water is a product of
formaldehyde-induced crosslinking phenomena in gelatin (17), the exclusion
of water from the SEGC gelatin shell with increasing concentrations of formaldehyde
in the PEG fill is a likely explanation for the PC 2 loadings with wavelengths
representing the NIR absorbance of water (1945 and 1420 run; Figure 4).
Thus, as the concentration of formaldehyde in the SEGC fill material increased,
the NIR spectra of the empty SEGCs, verified by decreases in specific spectral
regions (1420 and 1945 nm; Figure 4) confirmed loss of water in the gelatin
shell. Because the first PC in any PCR model represents most of the variation
of the original data set, reevaluation of the graph depicting loadings for
the first PC (Figure 3) proved useful. The loadings of the first PC actually
describe a baseline shift in the spectra that arises from a change in water
concentration. The absorbance spectrum of water in the NIR region is so
intense that a small change in water concentration affects the entire spectral
baseline. In the third graph of PC loadings (Figure
5), the presence of positive wavelength weighting deviations (which
are not due to water absorbance) at 1780 and 2200 nm confirms the hypothesis
that during the formaldehyde-induced crosslinking of gelatin, there are
chemical bonds broken and/or formed (1,18). These crosslink bonds absorb
vibrational energy and give rise to the harmonic and overtone frequencies
in the NIR region.
Gelatin crosslinking, initiated by formaldehyde introduction into the
PEG fill of a SEGC, was detected in the gelatin shell of the latter using
NIR spectrophotometry. When NIR was coupled to principal component analysis,
it was possible to predict, by taking the NIR spectra of empty SEGCs, the
amount of formaldehyde placed in the original fill material. Although not
shown in Figure 2, it is likely that with significantly
higher (and less pharmaceutically relevant) concentrations of formaldehyde
in the fill material of SEGCs, the reactive groups (E-amino and guanidino
functionalities of lysine and arginine, respectively) in the gelatin molecule
would have been saturated. In this connection, NIR would not have been as
amenable to predicting the amount of formaldehyde originally encapsulated
by the SEGC, since much of the formaldehyde would remain unreacted in its
PEG fill.
By combining NIR spectrophotometry with PCR, the determination of specific
contribution by wavelength to overall N!R spectral variation proved facile.
Water content of the SEGC, like our previous work with hard gelatin capsules
(10,13), proved to be the largest determinant in the variation of capsule
spectra representing different amounts of formaldehyde in the PEG fill.
With increasing concentration of formaldehyde in the PEG fill of the SEGC,
NIR spectrophotometry was able to distinguish a concomitant decrease of
water in the gelatin shell of the same capsule.
These laboratories are currently investigating computer-assisted molecular
modeling, which may permit not only the rapid, nondestructive analysis of
intact capsules with or without drug inside, but also the connection of
NIR to fundamental molecular motions. The latter information would verify
proposed gelatin crosslinking mechanisms.
REFERENCES
1. G. A. Digenis, T. B. Gold and V. P. Shah. Cross-linking of gelatin
capsules and its relevance to their in vitro-in vivo performance. J.
Pharm. Sci., 83:915-921(1994).
2. H. Fraenkel-Conrat, M. Cooper and H. S. Olcott. Reaction of formaldehyde
with proteins. J. Am. Chem. Soc., 67:950-954 (1945).
3. P. Davis and B. E. Tabor. Kinetic study of the crosslinking of gelatin
by formaldehyde and glyoxal. J. Polym. Sci., A1:799-815 (1963).
4. S. K. Taylor F. Davidson and D. W. Ovenall. Carbon-13 nuclear magnetic
resonance studies on gelatin crosslinking by formaldehyde. Photogr.
Sci. Eng., 22:134-138(1978).
5. K. Albert, B. Peters, E. Bayer, U. Treiber and M. Zwilling. Crosslinking
of gelatin with formaldehyde; a 13-C NMR study. Z. Naturforsch., 41b:35
1-358 (1986).
6. T. B. Gold, S. L. Smith and G. A. Digenis. Studies on the influence
of pH and pancreatin on 13C-formaldehyde-induced gelatin cross-links
using nuclear magnetic resonance. Pharm. Dev. Tech., 1:21-26 (1996).
7. G. A. Digenis and T. B. Gold. Chemistry of gelatin cross-linking.
Pharm. Res., 11:S146(1994).
8. J. T. Carstensen and C. T. Rhodes. Pellicle formation in gelatin capsules. Drug Dev. Ind. Pharm., 19:2709-2712 (1993).
9. C. B. Bottom, M. Clark and J. T. Carstensen. Dissolution testing of
soft shell capsules - acetaminophen and nifedipine. J. Pharm. Sci.,
in press.
10. T. B. Gold, R. G. Buice, Jr., R. A. Lodder and G. A. Digenis. Determination
of extent of formaldehyde-induced cross-linking and moisture content of
intact hard gelatin capsules by near-infrared spectrophotometry. Pharm.
Res., 12:S294 (1995).
11. E. Doelker and A. C. Vial-Bernasconi. Interactions contenant-contenu
au sein des capsules gelatineuses et evaluation critique de leurs effects
sur 1a disponibilite des principes actifs, S. T. P. Pharma., 4:298-306
(1988).
12. H. Mohamad, R. Renoux, S. Aiachc, and J. M. Aiache. Study on the
biopharmaceutical stability of medicines: application to tetracycline hydrochloride
capsules I. In vitro study. S. T. P. Pharma., 2:531-535 (1986).
13. T. B. Gold, R. G. Buice, Jr., R. A. Lodder and G. A. Digenis, Pharm.
Res., in press.
14. R. G. Buice, Jr., T. B. Gold, R. A. Lodder and G. A. Digenis. Determination
of moisture in intact gelatin capsules by near-infrared spectrophotometry.
Pharm. Res., 12:161-163 (1995).
15. J. E. Jackson. A User's Guide to Principal Components, John
Wiley & Sons, New York, 1991, pp:3-30.
16. R. A. Lodder, M. Selby, and G. M. Hieftje. Dejection of capsule tampering
by near-infrared reflectance analysis. Anal. Chem., 59:1921-1930
(1987).
17. I. D. Robinson. Rate of crosslinking of gelatin in aqueous solution.
J. Appl. Polym. Sci., 8:1903-1918 (1964).
18. R. F. Goddi and D. A. Delker. Spectra-structure correlations for
the near-infrared region. Anal. Chem., 32:140-141(1960).
Figure and Table Captions:
Table I: Solutions injected into the soft gelatin capsules. HCHO = formaldehyde,
PEG = polyethylene glycol
Figure 1. NIR spectra of empty soft gelatin capsules,
which had been filled with polyethylene glycol +/- formaldehyde solution.
Figure 2. Correlation of actual concentrations
of 37% formaldehyde solutions/ml PEG, encapsulated for 48 hours in soft
gelatin capsules, to concentrations predicted by NIR spectrophotometry of
empty, dry capsules(r2 = 0.988; passed cross-validation, SEE=O.024
v/v%, SEP=0.040 v/v%; p<0.05, f test). Nine of the thirteen generated
principal components were utilized in the prediction model. The calibration
line, representing r2=1, is shown superimposed on the data points.
Figure 3. Loadings for the first principal component,
showing wavelengths weighted by PCR in the NW spectral changes of empty
soft gelatin capsules (filled for 48 h with polyethylene glycol I formaldehyde
solution, then emptied and dried).
Figure 4. Loadings for the second principal component,
showing wavelengths weighted by PCR in the NIR spectral changes of empty
soft gelatin capsules (filled for 48 h with polyethylene glycol I formaldehyde
solution, then emptied and dried).
Figure 5. Loadings for the third principal component, showing wavelengths weighted by PCR in the NIR spectral changes of empty soft gelatin capsules (filled for 48 h with polyethylene glycol I formaldehyde solution, then emptied and dried).
