This work used dispersive Raman spectroscopy to evaluate acetaminophen in commercially available formulations as an analytical methodology for quality control in the pharmaceutical industry. Raman spectra were collected using a near-infrared dispersive Raman spectrometer (830 nm, 50 mW, 20 s exposure time) coupled to a fiber optic probe. Solutions of acetaminophen diluted in excipient (70 to 120% of the commercial concentration of 200 mg/mL) were used to develop a calibration model based on partial least squares (PLSs) applied to Raman spectra of solutions and, subsequently, obtain linearity, accuracy, precision (repeatability), and sensitivity of the method using the near-infrared spectroscopy (NIRS) as a gold standard method. This model was used to predict the acetaminophen concentration in commercial samples from different lots of acetaminophen formulations (200 mg/mL) with a PLS-prediction error of about 0.6%. Commercial medicines had PLS predicted concentrations errors below 2.5%, whereas NIRS had an error of about 3.7% compared to the label concentration. It has been demonstrated the applicability of Raman spectroscopy with fiber probe for quality control in pharmaceutical industry of commercial formulations.
The acetaminophen is a well-known analgesic and antipyretic that has been widely used because of the low side effects and still being well tolerated when administered in therapeutic quantities. Routine techniques for measuring acetaminophen include chromatographic methods (high performance liquid chromatography—HPLC), spectrophotometric methods (ultraviolet/visible, near-infrared and infrared spectroscopy, including Fourier-transform near-infrared spectroscopy (NIRS), and fluorescence spectroscopy), flow-injection spectroscopy, chemiluminescence methods, capillary electrophoretic methods, and electroanalytical methods [
The classical techniques for pharmaceutical drugs evaluation, such as chromatography, are sensitive and very specific, but these techniques are time consuming and of high costs, needing samples to undergo a preprocessing prior analysis, which may ultimately destroy the sample. To address this problem, optical methods have been proposed for the quantitative determination of active pharmaceutical ingredients (APIs) in drugs and medicines, most of them based on absorption and reflectance spectrometric techniques in the ultraviolet-visible and infrared regions [
Vibrational spectroscopy such as NIRS and Raman spectroscopy have advantages which lie on spectral selectivity and representativity and have been proposed as a quality control tool. The NIRS has grown substantially as a quality control technique in the pharmaceutical industry, including analysis of acetaminophen [
Raman spectroscopy has been used in the quantitative analysis of API, noninvasively and nondestructively, with advantages such as the need for small amount of sample, virtually no sample preparation, no sample destruction, and short time to perform the analysis [
Chemometric techniques such as partial least squares (PLSs) have been used in the analytical chemistry in order to predict the amount of a specific compound in a sample using the spectral information [
The process analytical technology (PAT) aims at the development of strategies to the control of the production processes in the pharmaceutical industry [
The purpose of this study was to assess the feasibility of a quantitative evaluation of acetaminophen (commercial drop’s presentation) based on dispersive Raman spectroscopy using fiber optic probe and multivariate statistics PLS (RS/PLS quantitative model) and to verify selectivity, linearity, range, precision, and accuracy, in order to check the ability of Raman spectra and PLS to be used as an analytical technique to attend industries requirements and compare with the NIRS technique already used for this. The developed model was then employed to predict the concentration of acetaminophen in samples of commercial medicines containing this API with 200 mg/mL (drop’s), aiming at its use as an analytical technique of choice. The results were compared to the ones obtained by NIRS, assigned by the manufacturer of the drugs.
Solutions of nine different concentrations of acetaminophen were prepared by the laboratory of a leading manufacturer, representing the range from 70 to 130% of the concentration found in the commercially available drug, diluted in the same excipient used in the commercial drug (Table
Set of acetaminophen solutions, respective nominal concentrations, and solvents used in the dilution.
Sample | API nominal concentration (mg/mL) | API nominal concentration (%) | Solvent |
---|---|---|---|
J | 0 | 0 | Excipient |
D | 140 | 70 | Excipient |
H | 160 | 80 | Excipient |
C | 180 | 90 | Excipient |
B | 190 | 95 | Excipient |
G | 200 | 100 | Excipient |
PEG | 200 | 100 | Polyethylene glycol |
A | 210 | 105 | Excipient |
I | 220 | 110 | Excipient |
E | 240 | 120 | Excipient |
F* | 260 | 130 | Excipient |
The solution samples and commercial medicines were submitted to the near-infrared dispersive Raman spectroscopy technique using the Raman spectrometer detailed elsewhere [
Schematic diagram of the dispersive Raman spectrometer used in the experiment. Fiber cable has a “six around one” configuration as described in De Lima et al. [
The procedures of spectra calibration, fluorescence background subtraction, cosmic rays removal, and normalization were performed using the software OriginPro 7.5 (OriginLab Corp., MA, USA). The Raman shift calibration was done by collecting the spectrum of naphthalene and correlating the spectrum band positions (pixel) with the known band shift (cm−1) with a 3rd-order polynomial fitting. To remove the fluorescence background, it was performed the baseline correction, in which straight lines were fitted in the spectra valleys and subtracted from the gross spectrum, remaining the high frequency Raman bands. Spectra were then normalized using the band of polyethylene glycol (main constituent of the excipient) at 1450 cm−1 and plotted in the spectral range of 600 to 1800 cm−1.
The PLS uses the nominal concentration information to obtain factors or latent variables, which are correlated to the principal components vectors, calculated by the PCA (principal component analysis) method [
In the PLS model, the Raman spectra and the nominal concentrations were considered as independent (ordinates) and dependent (abscissas) variables, respectively, and, under Matlab 4.2 and PLS Toolbox 1.5, it was used the command
In order to define the number of lv to be used in the model and to predict the concentrations of the dilutions and drug samples with the least error of prediction, it was established a cross-validation procedure, using the Raman spectra (independent variables) from the replicate dataset, with the number of lv changing from 1 to 5, through the function
Standard error of prediction (SEP) of the PLS model as a function of the lv = 1 to 5 calculated using a replicate dataset.
The spectra of dilutions and commercial drugs were then used in the PLS model as new independent variables using the number of lv = 1 and the predicted concentrations were then calculated.
The predicted concentration of the replicate samples were used to estimate the parameters [
In this study, it was defined the range of 70% to 120% (140 to 240 mg/mL), due to problems of solubility of the API at 130%.
It is the ability of the method to measure a compound in the presence of other components, such as adjuvants. For this parameter, it was considered a solution of the acetaminophen with 200 mg/mL in pure excipient and in polyethylene glycol (6 replicates of each).
It is the ability of the method to demonstrate whether there is proportionality between the predicted and the nominal concentrations of the API in the sample, within a specific range. This parameter was determined by constructing a curve with solutions of five concentrations: 160, 180, 200, 220, and 240 mg/mL. We calculated the curve slope (
It is referred to the agreement between the results within a short period of time with the same analyst and same instrumentation. It is done by using at least nine determinations under the linear range of the method, that is, three concentrations (low, intermediate or 100%, and high) with three replications each. For the study, it was used the concentrations 160, 200, and 240 mg·mL−1 in triplicate, totaling nine solutions. The precision was expressed as relative standard deviation (RSD), according to the formula
It is the closeness of the results obtained by the method under study in relation to the true value (nominal). It is obtained by adding known amounts of the drug to the components of the excipient. It was determined from nine determinations in the range of the procedure, that is, concentrations of 160, 200, and 240 mg/mL in triplicates. The equation used to calculate the accuracy (AC) is
The Raman spectra of the acetaminophen solutions in the range of 600 to 1800 cm−1, with concentrations ranging from 140 to 240 mg/mL and normalized by the band at 1450 cm−1, are shown in Figure
Normalized Raman spectra of solutions of acetaminophen diluted in excipient (placebo) ranging from 140 to 240 mg/mL and used in the calibration dataset.
(a) Raman spectra of acetaminophen obtained indirectly by subtracting solution G (200 mg/mL in excipient) from solution J (only excipient) and (b) Raman spectra of acetaminophen adapted from Sigma-Aldrich home page (
By applying the PLS model in the Raman spectra of replicate samples (cross-validation) with selected number of lv, one could calculate the predicted concentration and the SEP that the model is subjected to. Figure
Real versus predicted concentrations of the RS/PLS method using the replicate dataset and lv = 1. SEP = 1.19 mg/mL (0.6%). The solid line represents the zero-error curve.
The results of the parameters obtained by RS/PLS are as follows.
The higher relative error of acetaminophen in standard excipient compared to acetaminophen in pure PEG was 2.4%, showing that the technique allows detection of the major component of the sample in different excipient (Table
Predicted concentrations of acetaminophen in different excipients and relative errors using the 200 mg/mL solution (G and PEG).
Sample | Predicted concentration using standard excipient (%) | Predicted concentration using polyethylene glycol (%) | Relative error (%) |
---|---|---|---|
1 | 99.1 | 98.9 | 0.2% |
2 | 100.3 | 98.7 | 1.6% |
3 | 100.2 | 98.4 | 1.8% |
4 | 100.1 | 98.3 | 1.8% |
5 | 99.7 | 98.2 | 1.5% |
6 | 100.6 | 98.2 | 2.4% |
It was determined the concentrations of five solutions (160, 180, 200, 220, and 240 mg/mL) and compared with the nominal concentration values. From the curve, it was obtained
It was used three concentrations with six replicate samples of the following concentrations: 160, 200, and 240 mg/mL. It was calculated the RSD for each concentration of RS/PLS compared to the nominal dilution (Table
Solutions of different concentrations (160, 200, and 240 mg/mL) used to calculate the precision (RSD) and accuracy (AC).
Real concentration (mg/mL) | VO | DAC (mg/mL) | SD (mg/mL) | RSD (%) | AD (mg/mL) | RD (%) | AC (%) |
160 | 168.5 | 167.7 | 1.0 | 0.6 | −7.7 | −4.6 | 104,8 |
166.8 | |||||||
168.1 | |||||||
167.1 | |||||||
169.2 | |||||||
166.7 | |||||||
200 | 194.9 | 197.8 | 2.9 | 1.5 | −2.2 | −1.1 | 99.0 |
198.2 | |||||||
198.2 | |||||||
195.6 | |||||||
196.4 | |||||||
203.1 | |||||||
240 | 223.6 | 224.8 | 13.4 | 6.0 | −9.9** | −4.3** | 98.5 |
232.5 | |||||||
233.2 | |||||||
*198.5 | |||||||
228.9 | |||||||
232.3 |
It was calculated by using the predicted concentrations of the 160, 200, and 240 mg/mL, in triplicate, being determined by means of RSD and recovery, as shown in Table
In this work, the RS/PLS was evaluated by parameters as linearity, accuracy, and precision in the same way of Orkoula et al. and Rossignoli et al. [
As the study of Szostak and Mazurek (2002) [
In order to achieve optimum yield and product quality, the global pharmaceutical industry is searching for better methods for tracking drugs throughout the production process, since most of the high-performance techniques currently used need sample preparation using solvents, are destructive and time consuming, and mostly need expensive inputs [
In complex systems, such as API in medicines and formulations, the classical univariate regression method may become unreliable because of possible band interactions and superimposition. The PLS method has been suitable for quantitative measurement in chemical analysis when applied to spectroscopic data [
According to the PAT guide, the gains in processes quality, safety, and/or efficiency can change depending on the product and these changes come from factors such as prevention of rejections, disposal, and reprocessing of products; possibility of carrying out analysis in real time with reduced or no sample preparation [
RS/PLS and NIRS values used to calculate the concentrations of acetaminophen in six commercial medicines and the errors of each method.
Commercial medicines | Nominal value (mg/mL) | RS/PLS concentration (mg/mL) | Error RS/PLS (mg/mL) | NIRS concentration (mg/mL) | Error NIRS (mg/mL) |
---|---|---|---|---|---|
Med_1 | 200,0 | 201,3 | 0,7 | 202,0 | 1,0 |
Med_2 | 200,0 | 202,7 | 1,4 | 204,0 | 2,0 |
Med_3 | 200,0 | 199,0 | 0,5 | 207,4 | 3,7 |
Med_4 | 200,0 | 201,5 | 0,8 | 202,0 | 1,0 |
Med_5 | 200,0 | 201,4 | 0,7 | 204,0 | 2,0 |
Med_6 | 200,0 | 203,8 | 1,9 | 206,4 | 3,2 |
This study demonstrated the effectiveness of the near-infrared dispersive Raman Spectroscopy with fiber optic probe as a quantitative method, with high linearity, selectivity < 2.5%, precision < 5%, and accuracy < 3%, with recovery values of <5% and SEP of 1.19 mg/mL (0.6%) for quantifying the acetaminophen diluted in excipient. For the quantification of API in the commercial formulation, the model showed concentration values as closer as 200 mg/mL, with less than 2.5% error compared to the labeled concentration, while, by NIRS method, the error found was 3.7%. These results suggest that RS/PLS with fiber optic probe could replace the NIRS method in quality control of the product under study, while maintaining the safety and efficacy.
V. G. Borio thanks CAPES/PROSUP for the Master Fellowship. L. Silveira Jr. and R. A. Nicolau thank CNPq for the Productivity Fellowship (process no. 305610/2008-2 and 314455/2009-4, respectively). H. P. M. de Oliveira thanks FAPESP (2006/56701-3) and CNPq (479655/2008-1) for the financial support.