ABSTRACT
Three laboratories present a harmonised system for the retrospective comparison of south west Asian heroin. It consists of an improved gas chromatographic (GC) profiling method and a computerised data retrieval. The investigations of the GC were necessary with a view to improve the reproducibility of the system. The necessity of a strict quality control is emphasized. The peaks of the GC profile were investigated for abundance, intensity, GC behaviour (reproducibility) and correlations; 16 of them were selected for describing the heroin profile in the database. The results from intra-lab profile comparisons are reported. The reproducibility of the analysis was good and the variation between the samples was large, thus, allowing conclusions with a high degree of certainty. The criteria of similarity were defined. The system is successfully running in all three labs. In connection with inter-laboratory comparison, the aspects of method harmonisation and standardisation are discussed. It appeared that the GC method is a very subtile one, urging for a strict standardisation between the three labs. Despite a long cooperation between three well-equipped and experienced labs, a more or less serious loss of reproducibility was noticed in the inter-lab results in comparison with the intra-lab results. The loss could for the greater part be attributed to the (limits of the) GC technique; a number of compounds, necessary for making the discrimination between samples, showed difficult chromatographic behaviour, leading to insufficient inter-lab reproducibility. Using the actual variables, improvements in performance can hardly be expected in the near future. The loss of reproducibilty implies that the number of false positive matches in a database search increases. This may strongly reduce the value of a relatively large, international database. The study shows that so far, the best option for international comparison is the analysis in a central laboratory. The idea of local determination at a large number of national labs and the use of a common database is not a realistic aim for this type of analysis.
