Cannabis profiling of seized samples: An intra-location variability study using gas chromatography-mass spectrometry profiles and multivariate data analysis.
Drug Test Anal
; 2023 Sep 11.
Article
em En
| MEDLINE
| ID: mdl-37691538
ABSTRACT
Yearly, cannabis belongs to the most seized drugs worldwide. During judicial investigations, illicit cannabis profiling can be performed to compare seized herbal material. However, comparison is challenging because of the natural heterogeneity of the psychoactive crop. Gas chromatography-mass spectrometry (GC-MS) profiles, consisting of eight cannabinoids, were used to study the intra-location (within) and inter-location (between) variabilities. Decision thresholds were derived from the 95% and 99% confidence limits, applying Pearson correlation coefficients for the intra-location samples. The false negatives and false positives (FPs) determined the discriminative power of different pretreatments applied to obtain the lowest FP error rate possible. Initially, a 97 samples data set was used and with log transformation as pretreatment, a decrease in FPs from 38% and 45% FPs to 17% and 22%, for both confidence limits, respectively, was seen relative to internal standard normalization that was used as reference. An additional intra-plantation variability study with 130 samples verified whether the initial model contained sufficient within-location information, but this was not the case. Hence, a combined data matrix was constructed with all seized samples. Log transformation provided the best FP results for both limits, that is, an improvement from 58% and 64% to 21% and 26%, respectively, was seen. The representativeness of these 'linked' thresholds was demonstrated using both cross-validation and an external set, for which similar FP results as for the calibration set were obtained. By applying data pretreatment, a significant improvement was observed to distinguish seized samples. However, the FP rate is still not at an acceptable level to defend in court.
Texto completo:
1
Base de dados:
MEDLINE
Idioma:
En
Ano de publicação:
2023
Tipo de documento:
Article