Cocaine is the second most commonly seized drugs of abuse entering Belgium, both for local consumption and for distribution around Europe. A challenging issue when analysing cocaine samples is the presence of many different adulterants and/or cutting agents to increase drug volumes and dealer profits. Analyses of cocaine samples are routinely performed with conventional techniques: colour testing and/or infrared spectrometry for screening and chromatographic techniques for identification and quantification.
The aim of this study was to develop a technique to identify and quantify cocaine in seized drug powders using Attenuated Total Reflectance Fourier Transform Infrared (ATR-FTIR) spectrometry and multivariate analysis methods. ATR-FTIR spectrometry combined with chemometrics is a low-cost and less time-consuming technique in comparison with chromatographic techniques.
Circa two hundred cocaine powders were analysed using a mobile ATR-FTIR spectrometer, gas chromatography-mass spectrometer and gas chromatography-flame ionization detector. The percentage of cocaine in the samples ranged from 18% to 99%. Levamisole, phenacetin, diltiazem, caffeine, hydroxyzine, boric acid, benzocaine and lidocaine were found as byproducts. Chemometric analysis was performed on the collected spectral data using the Analyze IQ Limited software package (Version 3, Galway, Ireland). Several pre-processing techniques were tested followed by different chemometrical methods to identify and quantify the target cocaine. The most promising models for identification were the Support Vector Machine with Weighted Spectral Linear Kernel (SVMW) and the Spectral Attribute Voting technique with both an average error of 0.5%. For the quantification, Principal Component Regression and SVMW were superior with an error in prediction of respectively 11.2% and 8.9%.
Based on the results we can conclude that these models can be used to identify and quantify a wide variety of cocaine mixtures in routine laboratory analyses but also during on-site analysis by law enforcement. Moreover, the use of these models allows a high throughput, low-cost and quick identification and quantification in comparison to the conventional chromatographic techniques.