This article is the first of a three part series on quantitative prediction using Raman spectroscopy.
Drug eluding stents (DES) are an important therapy for coronary patients. Stents are metal wire meshes used to open clogged arteries. The coating of the stent contains a powerful immunosuppressant drug that prevents clotting of the artery near the stent due to the patient’s own immune system. The primary components of the stent coating are two polymers (PEVA and PBMA) and the drug, sirolimus. The Raman spectrum of the drug and the two polymers are shown in Figure 1.
Figure 1: Raman spectrum of drug (l. blue), PEVA (magenta), and PBMA (blue).
The Raman spectrum of the drug contains a strong peak at 1634 cm-1 due to the triene (three carbon carbon double bonds) functional group. Unique PBMA and PEVA bands occur at 600 and 630 cm-1 respectively. The CH region between 2800-3000 cm-1 shows clear peaks but the peaks are very broad and overlap.
A simple quantitative model for drug concentration was initially done using the peak amplitude at 1634 cm-1. This model was somewhat successful but a model with improved accuracy and robustness was desired. A significantly improved model was done using chemometric methods. The improved approach consisted of three major elements. A mixture design DOE was used to create a set of samples for the drug and polymer calibrations. The mixture design specified a set of samples where the concentration of the drug and both polymers were independently varied. The mixture design was supplemented with the use of production samples. The independent variation of all three major components allowed calibrations that are highly specific to the component of interest to be made.
Raman spectra of the production and mixture design experiments were then collected and the lab values for the drug concentration were measured on the same samples. The range of drug concentration was 0 to 37 %. (See Figure 2)
Figure 2: Drug calibration curve: x axis is laboratory data, y axis is predictions of PLS model. The production samples had a drug concentration of 28% (0.28).
The next critical step was appropriate spectral pre-processing before calibration development. A major source of variation in the Raman spectra is laser intensity. Laser intensity is the laser power divided by the laser spot size. In many practical cases, the laser spot size is hard to control due to small changes in the distance of the sample from the focusing lens. This is particularly important in this case were a short focal objective lens was used. The pre-processing method selected, standard normal variate (SNV), allows for the effective normalization of each Raman spectra. This normalization corrects for changes in the Raman signal strength with laser intensity. This led to a more accurate and robust calibration for all three major components, sirolimus, PEVA and PBMA.
After pre-processing a partial-least-squares (PLS) approach was used to build calibrations for the drug and both polymers. PLS regression uses spectral data to produce equations for constituent values of interest. PLS regression can be used when spectral peaks overlap and there are multiple sources of variation in the data set. PLS regression is the most common method for quantitative prediction in chemometrics. A future article in this series will discuss PLS methods in detail.
The next article in this series will summarize how the methods for drug and polymer concentration based on Raman spectroscopy were validated.
This article is a short summary of the following paper “Multivariate Analysis Applied to the Study of Spatial Distributions Found in Drug-Eluding Stent Coatings By Confocal Raman Microscopy” K. Balss, F. Long, V. Veslov, E. Akerman, G. Papandreou and C. Maryanoff, Anal. Chem., 80 , 4853-4859 (2008).