Detection capability of tetracyclines analysed by a fluorescence technique: comparison between bilinear and trilinear partial least squares models

According to the committee decision of 12 August 2002 (2002/657/EC) the capability of detection, CC(beta), must be set in all analytical methods not only at concentration levels close to zero but also at the maximum permitted limit (PL). In this work we describe a methodology which evaluates the capability of detection of a fluorescence technique with soft calibration models (bilinear and trilinear PLS) to determine tetracyclines (group B1 substances from annex 1 of Directive 96/23/EC). Its estimation is based on the generalisation of the procedure described in International Union of Pure and Applied Chemistry and in the ISO standard 11843 for univariate signals which evaluates the probabilities of false positive ((alpha)) and false negative ((beta)). The capability of detection, CC(beta), estimated from the secondorder signal and the trilinear PLS model is 9.93 (mu)g l^-1 of tetracycline, 17.75 (mu)g l^-1 of oxytetracycline and 26.31 (mu)g l^-1 of chlortetracycline, setting (alpha) and (beta) at 0.05. The capability of detection, CC(beta), determined around the PL (100 (mu)g kg^-1 in milk and muscle) with the second-order signal is 109.4 (mu)g l^-1 of tetracycline, 117.0 (mu)g l^-1 of oxytetracycline and 124.9 (mu)g l^-1 of chlortetracycline, setting (alpha) and (beta) at 0.05. The results were compared with those obtained with zero and first-order signals. The effect of the interferences on the capability of detection was also analysed as well as the number of standards used to build the models and their calibration range. When a tetracycline is quantified in presence of uncalibrated ones by means of the trilinear PLS model the errors oscillate between 14.70% for TC and 9.57% for OTC.