Development of a predictive model for batch membership of street samples of heroin

Street samples (n = 31) of heroin were analysed by gas chromatography with flame ionisation detection to determine opiate, noscapine and papaverine content. Using this data, the chromatograms obtained could be resolved into eight groups by visual examination of the data. The concentrations of opiates were significantly correlated (P < 0.05) with the exception of the pairs 6-O-monoacetylmorphine/noscapine and morphine/6-O-monoacetylmorphine. This precludes the use of simple cluster analysis for determining and predicting the relationship of different street samples. Application of Fisher's linear discriminant analysis to the data set indicated that 91.9% of the samples could be discriminated including pairs which could not be discriminated by eye. A blind trial (n = 2) resulted in the correct assignment to street sample. Application of such methods may provide, in the future, a powerful tool for the prediction of batch membership of drugs at the street level.