The design of a stability study is intended to establish a shelf-life based on testing a limited number of batches of a drug product, which is applicable to all future batches of the drug product manufactured under similar circumstances. In the FDA guidelines for stability testing of new drug products, the matrix design is suggested as the alternative statistical design to the full testing approach. In the last few years, the matrix design has received increasing attention from pharmaceutical companies in order to reduce the cost of stability studies. For this, the full and matrix stability test plans were used on three batches of human insulin pharmaceutical preparation stored under a kinetic mean temperature of 22.7 degrees C for 3 years. The statistical method was suggested by Ruberg and Stegeman (Biometrics, 1991, 47, 1059-1069) and Llabrés et al. (Int. J. Pharm. 2000, 204, 61-68), and was used for comparing batches, and the maximum allowable difference in slopes between batches was used as criteria for estimating a common shelf-life. The results obtained show the conservative nature of the matrix approach, although the difference between the matrix and full shelf-life was less than 1 month. The effect of the matrix method on shelf-life was also studied. The results showed that the model for estimating the shelf-life varied as a function of the permutation used, after testing for both slope and intercept poolability at alpha = 0.25. The batches should be pooled for both full testing and matrix design since the critical significance level required to obtain a power of 0.80 was achieved when the maximum allowable difference in slopes between batches was 1%/month, according to Ruberg and Stegeman. In contrast, when the proposed method by Llabrés et al. was used, the results varied in function of the maximum allowable difference in slopes.