The objective of the current study was to establish and evaluate a new technique to increase the accuracy of the in vitro / in vivo linear correlation of single and multiple dose ultra-sustained release theophylline (USRT) preparation (Xantium(R)) in hospitalized patients. In vitro dissolution data for theophylline were collected for 24 h using a USP I (basket) and USP 11 (paddle) methods. In vivo plasma concentration data were obtained from 8 patients after administration of either single or multiple doses of theophylline. Both in vitro and in vivo results were evaluated by zero-order, first-order, RRSBW, Hixson-Crowell, Higuchi, Hopfenberg, Langenbucher, modified Langenbucher and (Bt)(a) kinetic models. The individual linear correlations between each in vitro and in vivo percent results and their kinetic distributions were established and regression equations were obtained. The determination coefficient results of the linear kinetic correlations were found to be 0.994 and 0.997 for single and multiple doses by basket method and 0.992 and 0.998 for single and multiple doses by paddle method, respectively. Furthermore, the results of the linear correlations were found as 0.953 and 0.950 for single and multiple doses by basket method and 0.963 and 0.962 for single and multiple doses by paddle method respectively. Therefore, this study suggested that the accuracy of the linear correlation could be improved significantly by using linear kinetic correlation.