European Journal of Pharmaceutical Sciences 23 (2004) 355362

Development and validation of a direct, non-destructive quantitative

method for medroxyprogesterone acetate in a pharmaceutical
suspension using FT-Raman spectroscopy
T.R.M. De Beera , G.J. Vergoteb , W.R.G. Baeyensa, , J.P. Remonb ,
C. Vervaetb , F. Verpoortc
a Laboratory of Drug Quality Control, Department of Pharmaceutical Analysis, Ghent University, Harelbekestraat 72, B-9000 Ghent, Belgium
b Laboratory of Pharmaceutical Technology, Ghent University, Harelbekestraat 72, B-9000 Ghent, Belgium
c Department of Inorganic and Physical Chemistry, Ghent University, Krijgslaan 281 (S3), B-9000 Ghent, Belgium

Received 8 June 2004; received in revised form 17 August 2004; accepted 24 August 2004
Available online 2 November 2004

Abstract

A simple linear regression method was developed and statistically validated for the direct and non-destructive quantitative analysiswithout
sample preparationof the active pharmaceutical ingredient (API) medroxyprogesterone acetate (MPA) in an aqueous pharmaceutical sus-
pension (150 mg in 1.0 ml) using FT-Raman spectroscopy. The linear regression was modelled by plotting the highest peak intensity of the
vector normalized spectral band between 1630 and 1590 cm1 against different MPA standard suspension concentrations. At this band, no
spectral interferences from additives in the suspension are observed. The validated model was used for the quantification of a commercial sus-
pension (150 mg in 1.0 ml) of the commercialized preparations. The same standards and samples were used, respectively, for the development
and validation of a simple linear regression model and for the quantitative determination by means of HPLCwith sample preparationas
described for the related substances of MPA in the Ph. Eur. IV. The quantification results obtained by the FT-Raman method corresponded
with the claimed label concentration (150.01 0.96 mg/ml (n = 6)). Applying the HPLC method, however, a systematic error was observed
(157.77 0.94 mg/ml (n = 6)). The direct FT-Raman method hence appears the most reliable for the quantification of the MPA component in
suspension, compared to the HPLC method that requires sample preparation. The latter method provides a systematic error because the exact
volume or density of a suspension sample is unknown. A precise isolation of fixed volumes from a suspension is rather unfeasible because of
the continuous sagging of the suspended particles and their sticking to the used materials in the isolation process.
2004 Elsevier B.V. All rights reserved.

Keywords: FT-Raman spectroscopy; HPLC; Pharmaceutical suspension; Validation

1. Introduction vantages result from the fact that Raman spectroscopy is a

Since the last decade, high-technological improvements
in Raman spectrometers (McCreery, 2000; Long, 2002) cou-
pled with the important advantages offered by Raman spec-
troscopy, have made this spectroscopic technique more at-
tractive for the qualitative and quantitative determination of
pharmaceuticals (Vankeirsbilck et al., 2002). Substantial ad-
Corresponding author. Tel.: +32 926 48097; fax: +32 926 48196.
E-mail address: Willy.Baeyens@UGent.be (W.R.G. Baeyens).
fast and non-destructive technique and that spectra can be
recorded directly inside the packaging (glass, plastics and
blisters) without worth mentioning interferences. These ad-
vantages make it possible to measure APIs as well as addi-
tives as they appear in reality in drug formulations, without
any prior manipulation by sample treatment procedures. Un-
til now, mostly quantitative analyses of pharmaceutical solid
dosage forms (some inside their packaging) have been de-
scribed (Compton and Compton, 1991; Tsuchihashi et al.,
1997; Niemczyk et al., 1998; Skoulika and Georgiou, 2001,

0928-0987/$ see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejps.2004.08.009
356 T.R.M. De Beer et al. / European Journal of Pharmaceutical Sciences 23 (2004) 355362
2003; Yang and Irudayaraj, 2002; Vergote et al., 2002). The
non-destructive nature and speed of analyses performed on
solid materials make Raman spectroscopy an excellent and
promising process analytical technology (PAT) tool for phar-
maceutical production processes (blending, reaction moni-
toring, controlling the solid state of pharmaceutical crys-
tals during production processes, . . .) (Clegg and Everall,
2003; Vankeirsbilck et al., 2002; Christopher, 1999; Vergote
et al., 2004). Quantitative determination methods with Ra-
man spectroscopy of other pharmaceutical forms (creams,
suspensions, solutions, emulsions, etc.) are less frequent in
literature, so far.
Water is a very weak Raman scatterer and hence yields
no significant signal in Raman spectra. This is in contrast to
infrared (IR) spectroscopy where water provides a huge ab-
sorption signal, often overwhelming the signals of interest.
As a consequence, Raman measurements in aqueous envi-
ronments become most attractive (Gerrard and Birnie, 1992).
In the present work, a quantitative determination method by
FT-Raman spectroscopy of an aqueous pharmaceutical MPA
suspension was developed, validated and applied to com-
mercially available suspensions (Depo-Provera 150 mg in
1.0 ml). Beside the water advantage, it is of outmost impor-
tance to note that Raman spectroscopy allows direct measure-
ment of suspended API concentrations, without the need for
knowing the exact volume or density of the sample suspen-
sion. It is evident that different samples all provide different
volumes. Isolating exact volumes from a suspension is not
possible because of the continuous sagging of the suspended
particles and their sticking to the used materials in the iso-
lation process. This volume-sampling problem is confirmed
by HPLC analyses of the same samples, where the volume
claimed on the packaging of the samples (1.0 ml) was ac-
cepted as true value.
2. Materials and methods
2.1. Materials
Medroxyprogesterone acetate (MPA) was purchased
from Certa (Braine-l Alleud, Belgium). Methylparaben and
propylparaben, polyethylene glycol 4000, polysorbate 80 and
sodium chloride were provided by Alpha Pharma (Nazareth,
Belgium). Commercial MPA suspensions (Depo-Provera
150 mg in 1.0 ml, Pharmacia, Puurs, Belgium) were obtained
from a local drug store.
2.2. Standard and commercial suspensions
The drug content of six commercial suspensions (Depo-
Provera 150 mg in 1.0 ml) was first determined by FT-
Raman spectroscopy followed by HPLC analysis. Both meth-
ods were based on a simple linear regression model obtained
from a series of suspensions (standards) with known MPA
concentrations. The standards used for the development of
the HPLC calibration model were the same as those from
the FT-Raman model. Five standards were made ranging
from 100 to 200 mg/ml with increments of 25 mg/ml. They
were prepared by transferring a known amount of MPA in a
glass vial, followed by the addition of additive solution for
suspension till 1.00 ml. This additive solution contained ap-
proximately 28.9 mg of macrogol 4000, 2.41 mg of polysor-
bate 80, 8.68 mg of sodium chloride, 1.37 mg of methyl-
paraben, 0.150 mg of propylparaben and water ad injectabilia
per milliliter (www.rxlist.com/cgi/generic/medrox.htm).
The measurement conditions of the standards and the com-
mercial suspensions during FT-Raman spectrometry were al-
ways identical. As a consequence, all MPA measurements
had to be performed in the same type of vial. Therefore, the
Depo-Provera (150 mg in 1.0 ml) suspensions were directly
transferred from the syringe into the measurement vials.
2.3. Spectroscopic conditions
A Bruker FT spectrometer Equinox 55S (Bruker Optik, Et-
tlingen, Germany), equipped with the Raman module FRA
106 fitted to a cooled (77 K) germanium high sensitivity de-
tector D418-T, was used. The laser wavelength during the ex-
periments was the 1.064 m line from a diode laser pumped
Nd:YAG laser. All spectra were recorded at a resolution of
3 cm1 and a laser power of 1000 mW was used. Data col-
lection and data transfer were automated using the Bruker
OPUSTM software. A motorized positioner focused the laser
beam on the sample to obtain a maximum intensity of the
Raman signal. During the measurements, each suspension
vial (standards as well as samples) was fixed in a holder
to ensure that every suspension was measured in an iden-
tical position (Fig. 1). The recorded spectra were the results
of 1000 scan measurements. This number of scan measure-
Fig. 1. Sample chamber of the FT-Raman spectrometer, where a small mag-
netic stirrer was placed just below the round holder in which the vials were
fixed.
 T.R.M. De Beer et al. / European Journal of Pharmaceutical Sciences 23 (2004) 355362 357

ments corresponded to about 24 min of measuring, which is by the KolmogorovSmirnov test (Massart et al., 1997); (iii)
longer than the time the suspension remains homogeneous variation in the residuals is the same at all standard concentra-
after shaking it up (45 s). As a consequence, a small magnet tions (homoscedasticity), which can be tested by a Cochran
was introduced in the suspensions and stirred by a small mag- test (Massart et al., 1997; Miller, 1991; Cuadros Rodriguez
netic stirrer (12 mm 12 mm, Variomag , Oberschleiheim, et al., 1993). In order to apply unweighted linear least square
Germany), built in the Raman spectrometer right under the regression, suitability of the selected regression model was
suspension vial position (Fig. 1). The suspensions remained examined. Finally, the confidence and prediction intervals
homogeneous during Raman analysis as the maximum stir- as well for the estimated parameters of the selected regres-
ring speed (1000 rpm) during the measurements was used. At sion model as for the true line were calculated (Massart et al.,
this stirring speed, the 200 mg/ml standard could not be kept 1997). These statistical calculations for regression modelling
homogeneous, hence this standard was excluded for the FT- were done employing Statgraphics and Excel .
Raman calibration. A stronger magnetic stirrer with higher
stirring speed capacities was not applicable in the FT-Raman 2.6. Drug content determination of the commercial
spectrometer. suspensions
No internal standard was necessary and intersession cali-
bration was not done. The Raman signal was vector normal- From the validated regression calibration models (FT-
ized to compensate for any change in experimental conditions Raman and HPLC), the MPA content in six commercial sus-
(variation of the excitation intensity, sample positioning and pensions was estimated together with its confidence intervals
temperature) (Vergote et al., 2002). using Statgraphics . Each sample was measured six times.
2.4. HPLC method
3. Results and discussion
Parallel to the development of the FT-Raman calibration
model the same standards were used to develop the HPLC cal-
3.1. Development and validation of the FT-Raman
ibration model based on the assay for related substances of
simple linear regression calibration model
MPA described in the Ph. Eur. IV. The HPLC system (Merck
Hitachi, Tokyo, Japan) consisted of a pump (type L-7100), a
The FT-Raman spectra of the lab-made suspension and ad-
UV-detector (type L-7400), an integrator (type D-7000) and a
ditive solution (suspension without API) are shown in Fig. 2.
reversed-phase column (LiChrospher 100 RP C18 (5 m),
As the additive solution showed no signal at the spectral range
4 mm 25 cm). All measured suspensions (standards as well
of interest (17001550 cm1 ), the signal from the suspension
as samples) were transferred quantitatively in a volumetric
in this region must originate from MPA. As it is not com-
flask of 50.00 ml and dissolved with a mixture of tetrahydro-
pletely sure that the composition of the lab-made additive
furan and mobile phase (10:90, v/v). After dilution of this
solution is exactly the same as the commercial additive solu-
solution (1:100, v/v), an aliquot of 20 l was injected.
tion, also a spectrum of the latter was recorded. Therefore, a
commercial sample was centrifuged followed by the record-
2.5. Development and validation of the FT-Raman and
ing of a spectrum from the supernatans (commercial additive
HPLC simple linear regression calibration models
solution). Also, this spectrum showed no signal at the spectral
range of interest. The highest peak intensity of the vector nor-
The procedure for the development and validation of the
malized spectral band between 1630 and 1590 cm1 was the
simple linear regression calibration models was identical for
signal selected for the development of the calibration model
the FT-Raman method and the HPLC method. The mea-
and for the quantitative analysis of the commercial suspen-
sured signals (peak intensity of the vector normalized spec-
tral band between 1630 and 1590 cm1 for the FT-Raman
measurements and peak area for the HPLC measurements)
were plotted against the different MPA concentrations in the
standard suspensions. Each standard was measured six times.
By means of unweighted linear least square regression, the
parameters of a straight line through the plotted data were
estimated. This method involves three assumptions, to be
checked: (i) all the errors occur in the instrumental signal
measurements (the errors in making up the standards are neg-
ligible compared with the errors in measuring instrument sig-
nals); (ii) the signal measurement errors or residuals (differ-
ences between the measured signal and the signal predicted
by the calibration model) are normally distributed with mean
zero at every x-value (concentration), which can be tested Fig. 2. FT-Raman spectra of the suspension and the additive solution.
358 T.R.M. De Beer et al. / European Journal of Pharmaceutical Sciences 23 (2004) 355362

Table 1
The measured signals (six per standard suspension), the expected peak intensities calculated for every standard using the regression equation (expected values)
and the differences between the measured and expected peak intensities (residuals)
Standard suspension concentration (mg/ml) Measured peak intensities (AU) Expected peak intensities (AU) Residuals
99.9 0.10225 0.00039
0.10309 0.00045
0.10307 0.10264 0.00042
0.10334 0.00070
0.10239 0.00025
0.10267 0.00002
125.3 0.11015 0.00055
0.11087 0.00015
0.11077 0.11071 0.00006
0.10990 0.00080
0.11030 0.00040
0.11167 0.00096
150.2 0.11843 0.00019
0.11936 0.00074
0.11946 0.11862 0.00083
0.11755 0.00107
0.11781 0.00081
0.11739 0.00123
174.9 0.12717 0.00069
0.12587 0.00059
0.12580 0.12647 0.00066
0.12670 0.00023
0.12729 0.00082
0.12734 0.00087

KolmogorovSmirnov tests (Table 2), where the calculated

Fig. 3. Estimated simple linear regression curve through the measured FT-
Raman data (y = 0.00031772x + 0.0709038).
sions. Homogeneous sampling was assured as the suspen-
sions were stirred continuously (cf. 2.3).
After plotting the measured signals (six per stan-
dard; Table 1) against the different suspension concentra-
tions, the equation of the simple linear regression curve
through the data was estimated by unweighted linear
least square regression (Fig. 3). The regression equation
y = 0.00031772x+ 0.0709038 was found. The assumptions
concerning the application of the unweighted linear least
squares method were checked. (i) If no important sample
preparationsas in this caseare needed, the errors in mak-
ing up the standards can be considered negligible compared to
the errors in measuring the instrument signals (Miller, 1991).
(ii) The six residuals at every standard concentration (Table 1)
are normally distributed as proven mathematically by the
values for this test are smaller than the critical value (0.323)
at n = 6 and = 0.05. In addition, a t-test (Table 2) shows that
there is no significant difference of the means of the residuals
from zero for = 0.05 at every concentration level (p > 0.05).
Both tests allow us to conclude that the residuals are normally
distributed around zero at every concentration level. (iii) Ho-
moscedasticity of the normal distributed residuals is proven
mathematically by a Cochran-test (p = 0.40 > 0.05).
As the three assumptions concerning the residuals are met,
it is concluded that unweighted linear least squares regres-
sion can be applied. In a next step, it should be verified if
the chosen linear regression model adequately describes the
relationship between the standard concentration and the peak
intensity, within the examined concentration range (Massart
et al., 1997). Therefore, it is necessary to distinguish two char-
acteristics which are both described by the term linearity, but
which refer to different properties: on-line linearity (indicat-
ing the dispersion of the data around a calibration line) and
in-line linearity (indicating the absence of curvature on a cal-
ibration line). In-line linearity cannot be evaluated separately
from on-line linearity. The on-line linearity (LINOL (%)) can
be expressed as calculated from Eq. (1), where R.S.D.(b)
is the relative standard deviation of the slope (Cuadros
Rodriguez et al., 1996).
LINOL (%) = 100 [1 R.S.D.(b)] (1)
The maximum value for the relative standard deviation of
the slope is given by Eq. (2), where a t-value for a confidence
 T.R.M. De Beer et al. / European Journal of Pharmaceutical Sciences 23 (2004) 355362 359

Table 2
Results of the tests needed for checking the normal distribution of the residuals around zero for the FT-Raman simple linear regression calibration model
Concentration (mg/ml) KolmogorovSmirnov test ( = 0.05) t-test ( = 0.05)

Calculated value Critical value (n = 6) Average of the six residuals p-value

99.9 0.231 0.323 0.00016 0.90
125.3 0.186 0.323 0.00010 0.71
150.2 0.219 0.323 0.00029 0.47
174.9 0.248 0.323 0.00023 0.79

level of 1 ( = 0.05) with n2 degrees of freedom is usu-

ally used (n is the total number of measurements performed
for the creation of the calibration curve) (Cuadros Rodriguez
et al., 1993)
0.1
RSD (b) = (2)
t/2;n2
By combination of Eq. (2) and Eq. (1), the minimum on-line
linearity (linearity limit) can be calculated as expressed by
Eq. (3).


0.1
LINOL(min) (%) = 1 100 (3)
t/2;n2
When the calculated LINOL (%) is lower than LINOL(min)
(%), the dispersion of the data around the calibration line Fig. 5. Ninety-five percent prediction interval, 95% confidence interval and
is too high to apply adequately the ANOVA lack-of-fit test. 95% WorkingHotelling interval of the estimated FT-Raman calibration
The higher the random spread, the more difficult it is to model.
determine variance due to the chosen model. As the on-line
linearity (98.40%) is higher than the linearity limit (95.18%) to the data is not significant. From the observed data, the
for the suspension calibration curve, it was possible to apply fitted second-degree polynomial is given by the equation:
an ANOVA lack-of-fit test ( = 0.05) indicating the absence y = 0.0765056 + 0.00023264x 3.09688 107 x2 . Since
of curvature on the calibration curve. Since the calculated the p-value (0.18) for the t-test is higher than 0.05, the
p-value (0.36) for the lack-of-fit test is higher than 0.05, the quadratic regression coefficient can be considered as not
variance due to the random measurement errors is higher significant. The last step of the validation process is the
than the variance due to the error of the model, so that the calculation of the joint confidence intervals of the true
chosen linear model appears to be adequate for the observed parameters (slope, intercept and true response) of the model.
data. Another way to examine the linearity of the calibration In a linear regression model, slope and intercept are not
curve is proving by a t-test ( = 0.05) that the quadratic independently estimated and both parameters mutually in-
regression coefficient of a fitted second-degree polynomial fluence each other. Therefore, the use of a 95% ellipse joint
confidence region for slope and intercept (Fig. 4) is created,
which correlates the estimates for the slope (y-as) and the in-
tercept (x-as). All sets of slope and intercept that fall within

Fig. 6. Estimated simple linear regression curve through the measured HPLC
Fig. 4. Ninety-five percent joint confidence region for slope and intercept
data (y = 1.77807 107 x863.14).
of the estimated FT-Raman calibration equation.
360 T.R.M. De Beer et al. / European Journal of Pharmaceutical Sciences 23 (2004) 355362

Table 3 Table 5
The measured signals (six per standard suspension), the expected peak inten- Result of the test (Cochran test) needed to check the third assumption for
sities calculated for every standard using the regression equation (expected the HPLC simple linear regression calibration model and the results of the
values) and the differences between the measured and expected peak inten- suitability tests (ANOVA and t-test for the quadratic regression coefficient
sities (residuals) of the polynomial curve)
Concentration (mg/ml) Peak area Expected peak area Residuals Test ( = 0.05) p-value
358223 1338 Cochran-test 0.36
356852 33 ANOVA lack-of-fit test 0.11
0.02012 356057 356885 828 t-test for the quadratic regression co- 0.42
355900 985 efficient of the polynomial curve
357198 313
353288 3597
443838 184
448546 4892
0.025 449545 443654 5891
452631 8977
446097 2443
438852 4802
540003 1755
537597 651
0.03032 540005 538248 1757
524109 14139
534432 3816
538069 179
629177 4020
630801 2396
0.03566 635315 633197 2118
628858 4339
634374 1177
631505 1692 Fig. 7. Ninety-five percent joint confidence region for slope and intercept
of the estimated HPLC calibration model.
712425 993
709273 2159
0.04006 718453 711432 7021 limits the peak intensity will be for one measurement of a
711990 558 certain suspension concentration.
706824 4608
720282 8850
3.2. Development and validation of the HPLC simple
linear regression calibration model
the ellipse are considered to be included in the joint confi-
dence interval (Massart et al., 1997). The 95% confidence The development of the HPLC simple linear regression
interval from the observed data (Fig. 5) shows the limits of calibration model was done in the same way as the Raman
the true response (peak intensity) at a particular x-value (con- model. After the measured signals (six per standard; Table 3)
centration), which may be expected with 95% assurance. The were plotted against the different suspension concentrations,
95% WorkingHotelling confidence band (Fig. 5) describes the equation of the simple linear regression curve through
the limits of the true regression line with 95% assurance. In the data was estimated via the unweighted linear least square
fact, this WorkingHotelling band includes all possible re- method (Fig. 6): y = 1.77807 107 x 863.14. The tests to
gression lines of all possible combinations of the slope and check the assumptions for the application of the unweighted
intercept within the joint confidence interval. The 95 % pre- linear least square method are satisfying (Tables 4 and 5).
diction interval for measured peak intensities at known con- As the on-line linearity (99.33%) is higher than the linearity
centrations (Fig. 5) shows with 95% assurance within what limit (95.12%), it was allowed to use an ANOVA lack-of-fit

Table 4
Results of the tests needed for checking the normal distribution of the residuals around zero for the HPLC simple linear regression calibration model
Concentration (mg/ml) KolmogorovSmirnov test ( = 0.05) t-test ( = 0.05)

Calculated value Critical value (n = 6) Average of the six residuals p-value

0.02012 0.250 0.323 632 0.39
0.025 0.158 0.323 2931 0.19
0.03032 0.290 0.323 2546 0.34
0.03566 0.192 0.323 1525 0.21
0.04006 0.226 0.323 1776 0.44
 T.R.M. De Beer et al. / European Journal of Pharmaceutical Sciences 23 (2004) 355362 361

Fig. 8. Ninety-five percent prediction interval, 95% confidence interval and 95% WorkingHotelling interval of the estimated HPLC calibration model.

test ( = 0.05) for indicating the absence of curvature on the
calibration curve. This test and the significance t-test of the
quadratic regression coefficient from the polynomial curve
through the data (Table 5) show that the linear model seems to
be adequate. The confidence intervals of the true parameters
(slope, intercept and true response) are shown in Figs. 7 and 8.
3.3. Validation of the Raman and the HPLC method
The accuracy of the Raman and HPLC analyses were val-
idated by cross-referencing the actual and the calculated (us-
ing the calibration curves) MPA content after analysis of val-
idation samples containing known amounts of drug (125, 150
and 175 mg/ml) per suspension. The validation samples were
first analyzed by the Raman method followed by the HPLC
method. It must be taken into account that the calculated con-
centrations are subject to error for two reasons: (1) the errors
in the calibration line, and (2) the error in the measurement
itself (Miller, 1991). The calculated and actual values for
the validation samples are shown in Table 6 (every sample
was measured six times). There is a 95% probability that the
true concentration in the sample falls within the limits spec-
ified by the 95% confidence limits, which is the case. These
results confirm that the Raman and HPLC methods are ac-
curate analytical tools to determine the MPA content in the
suspensions.
3.4. Drug analysis of six commercial suspensions
(Depo-Provera 150 mg/1 ml)
The drug content calculations from the commercial sus-
pensions were performed identically as for the validation
samples. The MPA dosage per suspension calculated from
the FT-Raman data was significantly different from the re-

Table 6
Calculated and actual values for the validation samples
Actual (mg/ml) Raman HPLC

Calculated (mg/ml) 95% confidence limits (mg/ml) Calculated (mg/ml) 95% confidence limits (mg/ml)
124.9 126.5 [124.4; 128.6] 125.5 [124.3; 126.8]
150.0 148.7 [146.6; 150.8] 148.8 [147.6; 150.0]
175.2 176.6 [174.2; 179.1] 174.1 [172.9; 175.4]

Table 7
FT-Raman and HPLC analysis results of six commercial suspensions
Commercial sample Raman HPLC

Mean (mg/ml) (n = 6) 95% confidence limits (mg/ml) Mean (mg/ml) (n = 6) 95% confidence limits (mg/ml)
1 150.0 [147.9; 152.1] 158.5 [157.3; 159.7]
2 150.8 [148.7; 152.9] 157.1 [155.9; 158.3]
3 148.8 [146.7; 150.9] 156.3 [155.1; 157.5]
4 151.4 [149.3; 153.5] 157.7 [156.5; 158.9]
5 149.6 [147.5; 151.7] 158.8 [157.6; 160.0]
6 149.5 [147.4; 151.6] 158.3 [157.1; 159.5]
362 T.R.M. De Beer et al. / European Journal of Pharmaceutical Sciences 23 (2004) 355362
sults obtained by HPLC-analysis of the Depo-Provera sus-
pensions (average amount of MPA per suspension (n = 36):
150.01 0.94 mg/ml (Raman) versus 157.77 0.95 mg/ml
(HPLC) (p = 0.00002 < 0.05)) (Table 7). The latter method
will always provide a systematic error because the exact vol-
ume from a suspension is never known. A precise isolation
of fixed volumes from a suspension is rather impossible be-
cause of the continuous sagging of the suspended particles
and their sticking to the used materials in the isolation pro-
cess. To overcome this isolation and volume problem, the
density of a 50.0 ml self-made MPA suspension was deter-
mined ( = 1.0349 mg/ml). Using this density value, the ex-
act volume of six commercial samples was calculated after
weighting them precisely. On the basis of these volumes, their
drug contents were determined (n = 6): 152.3 0.85 mg/ml.
Also, these HPLC results were significantly different from
the FT-Raman results (p = 0.008). A possible reason for this
can be the errors caused by the sample preparation needed for
the HPLC measurements and the error on the density value.
Small changes of this value give relatively high changes of
the drug content values.
4. Conclusion
From the cited results, it can be concluded that the FT-
Raman spectroscopic method based on unweighted linear
regression calibration is most reliable for the quantitative
determination of the MPA component in aqueous suspen-
sion, compared to the HPLC method, which requires sample
preparation and the knowledge of the exact sample volumes
or density of the suspension.
References
Christopher, F.J., 1999. Review of pharmaceutical applications of Raman
spectroscopy. In: Pelletier, M.J. (Ed.), Analytical Applications of Ra-
man Spectroscopy. Blackwell Science, London, UK, pp. 224275.
Clegg, I., Everall, N., 2003. On-line measurement of crystalline forms.
Eur. Pharm. Rev. 3, 5662.
Compton, D.A.C., Compton, S.V., 1991. Examination of packaged
consumer-goods by using FT-Raman spectrometry. Appl. Spectrosc.
45 (10), 15871589.
Cuadros Rodriguez, L., Garcia Campana, A.M., Jimenez Linares, C.,
Roman Ceba, M., 1993. Estimation of performance characteristics of
an analytical method using the data set of the calibration experiment.
Anal. Lett. 26 (6), 12431258.
Cuadros Rodriguez, L., Garcia Campana, A.M., Bosque Sendra, J.M.,
1996. Statistical estimation of linear calibration range. Anal. Lett. 29
(7), 12311239.
Gerrard, D.L., Birnie, J., 1992. Raman spectroscopy. Anal. Chem. 64,
502R513R.
Long, D.A., 2002. The Raman Effect: a Unified Treatment of the Theory
of Raman Scattering by Molecules. Wiley-Interscience, Chichester,
West Sussex, England.
Massart, D.L., Vandeginste, B.G.M., Buydens, L.M.C., De Jong, S.,
Lewi, P.J., Smeyers-Verbeke, J., 1997. Handbook of Chemometrics
and Qualimetrics: Part A. Elsevier, Amsterdam, The Netherlands.
McCreery, R.L., 2000. Raman Spectroscopy for Chemical Analysis.
Wiley-Interscience, New York, USA.
Miller, J.N., 1991. Basis statistical methods for analytical chemistry, Part
2. Calibration and regression methods: a review. Analyst 116, 3
14.
Niemczyk, T.M., Delgado-Lopez, M.M., Allen, F.S., 1998. Quantitative
determination of Bucindolol concentration in intact gel capsules using
Raman spectroscopy. Anal. Chem. 70 (13), 27622765.
Skoulika, S.G., Georgiou, C.A., 2001. Rapid Quantitative determination
of ciprofloxacin in pharmaceuticals by use of solid-state FT-Raman
spectroscopy. Appl. Spectrosc. 55 (9), 12591265.
Skoulika, S.G., Georgiou, C.A., 2003. Rapid, noninvasive quantita-
tive determination of acyclovir in pharmaceutical solid dosage
forms through their poly(vinyl chloride) blister by solid-state fourier
transform Raman spectroscopy. Appl. Spectrosc. 57 (4), 407
412.
Tsuchihashi, H., Katagi, M., Nishikawa, M., Tatsuno, M., Nishioka, H.,
Nara, A., Nishio, E., Petty, C., 1997. Determination of metham-
phetamine and its related compounds using Fourier transform Raman
spectroscopy. Appl. Spectrosc. 51 (12), 17961799.
Vankeirsbilck, T., Vercauteren, A., Baeyens, W., Van der Weken, G., Ver-
poort, F., Vergote, G., Remon, J.P., 2002. Applications of Raman
spectroscopy in pharmaceutical analysis. TrAC: Trend. Anal. Chem.
21 (12), 869877.
Vergote, G.J., Vervaet, C., Remon, J.P., Haemers, T., Verpoort, F., 2002.
Near-infrared FT-Raman spectroscopy as a rapid analytical tool for the
determination of diltiazem hydrochloride in tablets. Eur. J. Pharm. Sci.
16, 6367.
Vergote, G.J., De Beer, T.R.M., Vervaet, C., Remon, J.P., Baeyens,
W.R.G., Diericx, N., Verpoort, F., 2004. In-line monitoring of a phar-
maceutical blending process using FT-Raman spectroscopy. Eur. J.
Pharm. Sci. 21, 479485.
Yang, H., Irudayaraj, J., 2002. Rapid determination of vitamin C by
NIR, MIR and FT-Raman techniques. J. Pharm. Pharmacol. 54 (9),
12471255.