Chocolate is rich in several bioactive compounds. The presence of these compounds increases the consumption of chocolate, due their biological activities. In this work a multivariate statistical techniques to test the concentrations of theobromine, catechin, epicatechin and caffeine in handmade chocolate samples with different cocoa content using the High Performance Liquid Chromatography (HPLC) were employed. To optimize the separation procedure a central composite design with three variables was used: initial percentage of methanol in mobile phase, gradient time and flow rate. The models generated were submitted to analysis of variance ANOVA (p<0,05) and it have no a significant lack of fit. To predict the optimal conditions, a multicriterial optimization methodology and surface response methodology were used. The best conditions obtained were: the initial percentage of methanol in 0 % (starting with 100 % of acetic acid solution 0.1 % v/v), changing to 100 % of methanol in 9.85 minutes, with flow rate of 0.74 mL min -1. These conditions brought the complete separation of peaks, with the resolution values equals to 1.3 for theobromine/catechin pairs, 1.8 for catechin/epicatechin, 1.3 for epicatechin/caffeine. The procedure optimized was validated and used to determine theobromine, catechin, epicatechin and caffeine into 10 samples of handmade chocolate with different cocoa contents, being obtained the range of concentration, in mg g-1, for the compounds: theobromine (1,56-13,9), catechin (0,018-0,607), epicatechin (0,048 – 1,67) and caffeine (0,235 – 1,51).