RESUMO
This study deals with the experimental and statistical comparison of six immunochemical techniques, including noninstrumental on-site and instrumental formats (EIA-EMIT and EZ-SCREEN; FPIA-ADx; RIA-Coat-A-Count; LI-Abuscreen ONTRAK; CBI-Triage), and three chromatographic techniques (TLC-Toxi-Lab; HPLC; HPLC-REMEDi drug profiling system), using GC-MS as a reference technique for analyzing amphetamines, barbiturates, benzodiazepines, cannabinoids, cocaine, methadone, and opiates in the urine of various kinds of drug users. The study reports (a) the values of sensitivity, specificity, false-positive rates, and false-negative rates of each technique; (b) the results of bayesian statistical analysis, which are based on prevalence values of the samples examined and expressed as positive and negative predictive values and cumulative predictive values for each single technique and for combinations of paired immunochemical and chromatographic techniques; and (c) the results of a rough classification of the various degrees of predictability of these techniques. Lastly, this study proposes a decision-making process for establishing the best combination of analytical techniques for the goals in question, according to the characteristics and facilities of each laboratory.