The construction of parallel editions of conventional tests for purposes of test security while maintaining score comparability has always been a recognized and difficult problem in psychometrics and test construction. The introduction of new modes of test construction, e.g., adaptive testing, changes the nature of the problem, but does not make it disappear. Items in adaptive testing pools may become overused and require replacement. However, in order to insure score comparability, important characteristics of the pool must remain constant. Three methods of selecting candidate new items and three methods of identifying items for replacement are developed and compared with each other and with a previous method through a simulation study. Results indicated that using conventional item statistics to screen items before deciding to seed them was important and effective in terms of maintaining the information structure of the adaptive test item pool. The online calibration of larger sets of seeded items from which to select replacements can substantially improve the ease with which the information structure of the pool can be maintained. (Contains 1 table, 11 figures, and 4 references.) (Author/SLD)