In the present study we evaluated the precision of the ELISA method to quantify caffeine in human plasma and compared the results with those obtained by gas chromatography. A total of 58 samples were analyzed by gas chromatography using a nitrogen-phosphorus detector and routine techniques. For the ELISA test, the samples were diluted to obtain a concentration corresponding to 50% of the absorbance of the standard curve. To determine whether the proximity between the I50 of the standard curve and that of the sample would bring about a more precise result, the samples were divided into three blocks according to the criterion of difference, in modulus, of the I50 of the standard curve and of the I50 of the sample. The samples were classified into three groups. The first was composed of 20 samples with I50 up to 1.5 ng/ml, the second consisted of 21 samples with I50 ranging from 1.51 to 3 ng/ml, and the third of 17 samples with I50 ranging from 3.01 to 13 ng/ml. The determination coefficient (R² = 0.999) showed that the data obtained by gas chromatography represented a reliable basis. The results obtained by ELISA were also reliable, with an estimated Pearson correlation coefficient of 0.82 between the two methods. This coefficient for the different groups (0.88, 0.79 and 0.49 for groups 1, 2 and 3, respectively) showed greater reliability for the test with dilutions closer to I50.